fix(models): show API models by default (#6089)

refactor(routes): move task domain into routes/task/ subpackage

.env.example

@@ -76,12 +76,24 @@ SEARXNG_INSTANCE=http://localhost:8080
 # Change this if another local service already uses 7000 (macOS AirPlay often does).
 # APP_PORT=7000
 
+# Optional HTTP address advertised in companion/mobile pairing codes. Set this
+# when Docker would otherwise advertise a container address or loopback. Use a
+# LAN or Tailscale IPv4 address, a single-label hostname, or an mDNS *.local
+# name that the phone can reach. HTTPS and public hostnames are not supported
+# by the current companion client. Do not include credentials, a path, query,
+# or fragment.
+# COMPANION_BASE_URL=http://192.168.1.50:7000
+
 # Development-only auth bypass for loopback requests.
 # Keep false for Docker, LAN, reverse proxy, and any shared deployment.
 # LOCALHOST_BYPASS=false
 
-# Mark session cookies Secure. Set true when Odysseus is served through HTTPS
-# by a trusted reverse proxy or private access gateway.
+# Mark session cookies Secure. Left unset, this follows the request scheme:
+# an HTTPS login gets a Secure cookie, a plain-HTTP one does not. Set true to
+# force it on, or false to force it off while you still serve plain HTTP.
+# Upgrading: this used to default to false. Drop a leftover SECURE_COOKIES=false
+# from your .env unless you still need that escape hatch — it keeps HTTPS logins
+# on a non-Secure cookie.
 # SECURE_COOKIES=true
 
 # Optional: pre-seed the first admin password during setup.
@@ -151,6 +163,21 @@ SEARXNG_INSTANCE=http://localhost:8080
 # Local HTTP setups may use the callback URL inferred by the application.
 # GOOGLE_OAUTH_REDIRECT_URI=https://your-domain.com/api/email/oauth/google/callback
 
+# Origin the MCP OAuth callback is sent back to, for remote (Streamable HTTP)
+# MCP servers that register it dynamically. Defaults to http://localhost:$APP_PORT,
+# which is right only when you reach Odysseus directly on that port. Set it for
+# HTTPS, reverse-proxy, hosted, and Docker installs — inside the container the
+# app always listens on 7000 and cannot see the host port map, so the default is
+# wrong there whenever APP_PORT is not 7000.
+#
+# Not for Google MCP servers. Those use Desktop App credentials, and Google only
+# accepts loopback redirect URIs for that client type, so a public origin here is
+# rejected with redirect_uri_mismatch. Leave it unset for a Google-only install:
+# the loopback default is what Google wants, and remote users finish through the
+# paste-back page, which never has to load the redirect.
+# https://developers.google.com/identity/protocols/oauth2/native-app
+# OAUTH_REDIRECT_BASE_URL=https://your-domain.com
+
 # ============================================================
 # Misc
 # ============================================================

.gitattributes

@@ -15,6 +15,13 @@ docker/entrypoint.sh text eol=lf
 *.cmd text eol=crlf
 *.bat text eol=crlf
 
+# Vendored third-party bundles in static/lib/ are published minified artifacts
+# and must stay byte-identical to what npm ships — stripping trailing whitespace
+# to satisfy `git diff --check` would desync them from the upstream release. Turn
+# the whitespace check off for that tree instead, and keep the bundles out of
+# GitHub's language statistics.
+static/lib/** -whitespace linguist-vendored
+
 # Binary assets — never normalize.
 *.png  binary
 *.jpg  binary

.github/ISSUE_TEMPLATE/bug_report.yml

@@ -26,6 +26,18 @@ body:
         - label: I am running the latest code from the `dev` branch (the default branch you get on clone, where fixes land first) and the bug still reproduces there. Please `git pull` the latest `dev` before filing.
           required: true
 
+  - type: input
+    id: revision
+    attributes:
+      label: Odysseus Revision
+      description: |
+        From the repository root (on the host when using Docker), run
+        `git show -s --abbrev=12 --format='%h (%cs)' HEAD`
+        and paste the output exactly.
+      placeholder: "1fef4929cf1d (2026-08-11)"
+    validations:
+      required: true
+
   - type: dropdown
     id: install-method
     attributes:

.github/pull_request_template.md

@@ -28,6 +28,7 @@ Fixes #
 - [ ] This PR targets `dev`
 - [ ] My changes are limited to the scope described above — no unrelated refactors or whitespace changes mixed in.
 - [ ] I actually ran the app (`docker compose up` or `uvicorn app:app`) and verified the change works end-to-end. Type-checks and unit tests are not enough.
+- [ ] I did not run the app/runtime validation and stated that gap in **How to Test**. Leave this unchecked when the app-run box above is checked.
 
 ## How to Test
 

.github/scripts/check-issue-description.js

@@ -41,6 +41,14 @@ module.exports = async ({ github, context, core }) => {
       break;
 
     case 'bug': {
+      const revisionText = section('Odysseus Revision');
+      if (!/^[0-9a-f]{12} \(\d{4}-\d{2}-\d{2}\)$/i.test(revisionText)) {
+        failures.push(
+          '**Odysseus Revision** — paste the 12-character commit SHA and date, ' +
+          'for example `1fef4929cf1d (2026-08-11)`',
+        );
+      }
+
       if (!section('Install Method')) {
         failures.push('**Install Method** — select how you installed Odysseus');
       }

.github/scripts/check-pr-description.js

@@ -21,11 +21,11 @@ module.exports = async ({ github, context, core }) => {
     return strip(m?.[0].replace(new RegExp(`#+\\s+${heading}`, 'i'), '') ?? '');
   }
 
-  const problems = [];
+  const descriptionProblems = [];
 
   // 1. Summary must be filled in.
   if (section('Summary').length < 20) {
-    problems.push('**Summary** is empty or too short — describe what changed and why.');
+    descriptionProblems.push('**Summary** is empty or too short — describe what changed and why.');
   }
 
   // 2. Linked Issue must reference a real issue. Accept a bare #NNN, a closing
@@ -34,18 +34,18 @@ module.exports = async ({ github, context, core }) => {
   const linkedSection = section('Linked Issue');
   const hasIssueRef = /#\d+\b/.test(linkedSection) || /\/issues\/\d+/.test(linkedSection);
   if (!linkedSection || !hasIssueRef) {
-    problems.push('**Linked Issue** — add a reference like `Fixes #NNN`, a bare `#NNN`, or a link to the issue.');
+    descriptionProblems.push('**Linked Issue** — add a reference like `Fixes #NNN`, a bare `#NNN`, or a link to the issue.');
   }
 
   // 3. At least one Type of Change box must be checked.
   const typeBlock = body.match(/##\s+Type of Change[\s\S]*?(?=\n##\s|$)/i)?.[0] ?? '';
   if (!/- \[x\]/i.test(typeBlock)) {
-    problems.push('**Type of Change** — check at least one box.');
+    descriptionProblems.push('**Type of Change** — check at least one box.');
   }
 
   // 4. Duplicate-search checklist item must be checked.
   if (!/- \[x\] I searched/i.test(body)) {
-    problems.push('**Checklist** — check the duplicate-search box to confirm you searched existing issues and PRs.');
+    descriptionProblems.push('**Checklist** — check the duplicate-search box to confirm you searched existing issues and PRs.');
   }
 
   // 5. How to Test must contain enough real detail for a reviewer to act on.
@@ -53,7 +53,83 @@ module.exports = async ({ github, context, core }) => {
   //    code block — so we only require non-trivial content, not a specific shape.
   const howTo = section('How to Test');
   if (howTo.length < 30) {
-    problems.push('**How to Test** — explain how a reviewer can verify this change. Numbered steps, the commands you ran, or a short code block all work — give a sentence or two of real detail (not just "tested locally").');
+    descriptionProblems.push('**How to Test** — explain how a reviewer can verify this change. Numbered steps, the commands you ran, or a short code block all work — give a sentence or two of real detail (not just "tested locally").');
+  }
+
+  // Classify paths from GitHub's API. This workflow runs in the privileged base
+  // context, so it must never check out or execute code from the PR branch.
+  const changedFiles = await github.paginate(github.rest.pulls.listFiles, {
+    owner, repo, pull_number: prNum, per_page: 100,
+  });
+  const changedPaths = changedFiles.map(file => file.filename);
+
+  function isUiSensitivePath(filename) {
+    const path = filename.toLowerCase();
+    return path.startsWith('static/')
+      || path.startsWith('templates/')
+      || /\.(?:html?|css|svg)$/.test(path);
+  }
+
+  function isDocsOnlyPath(filename) {
+    const path = filename.toLowerCase();
+    return /\.(?:md|mdx|rst|adoc|txt)$/.test(path)
+      || (path.startsWith('docs/') && !isUiSensitivePath(path));
+  }
+
+  function isRuntimeSensitivePath(filename) {
+    const path = filename.toLowerCase();
+    if (isUiSensitivePath(path)) return false;
+    if (path.startsWith('tests/') || path.startsWith('.github/')) return false;
+    return /^(?:app\.py|routes\/|services\/|src\/|core\/|mcp_servers\/|scripts\/|docker\/)/.test(path)
+      || /^(?:dockerfile|docker-compose.*\.ya?ml|requirements(?:-optional)?\.txt|pyproject\.toml|setup\.py)$/.test(path)
+      || /\.(?:py|sh|ps1|bat)$/.test(path);
+  }
+
+  let classification = 'tooling';
+  if (changedPaths.some(isUiSensitivePath)) {
+    classification = 'UI-sensitive';
+  } else if (changedPaths.some(isRuntimeSensitivePath)) {
+    classification = 'backend/runtime';
+  } else if (changedPaths.length > 0 && changedPaths.every(isDocsOnlyPath)) {
+    classification = 'docs-only';
+  }
+
+  const appRan = /- \[x\]\s+I actually ran the app\b/i.test(body);
+  const appNotRun = /- \[x\]\s+I did not run the app\/runtime validation\b/i.test(body);
+  // Anchor on the wording, not the template's emphasis: a ticked box the author
+  // retyped without the surrounding ** renders identically on the PR page, so
+  // treating it as unchecked is invisible from their side. Matches the two
+  // attestations above, which already ignore formatting.
+  const screenshotChecked = /- \[x\]\s+[*_]{0,2}Screenshot or short clip[*_]{0,2}/i.test(body);
+  const screenshotSection = section('Screenshots / clips');
+  const hasVisualEvidence = /!\[[^\]]*\]\([^)]+\)|<(?:img|video|source)\b[^>]*(?:src|href)=|https?:\/\/[^\s)]+/i.test(screenshotSection);
+  const evidenceGaps = [];
+  let needsRuntimeValidation = false;
+  let needsVisualEvidence = false;
+
+  if (classification === 'backend/runtime' || classification === 'UI-sensitive') {
+    if (appRan && appNotRun) {
+      needsRuntimeValidation = true;
+      evidenceGaps.push('The app-run and explicit not-run boxes are both checked. Select the one state that is true.');
+    } else if (!appRan) {
+      needsRuntimeValidation = true;
+      if (appNotRun) {
+        evidenceGaps.push('The author explicitly reports that app/runtime validation was not performed.');
+      } else {
+        evidenceGaps.push('App/runtime validation is not author-attested. Check the run box only after running it, or check the explicit not-run box and describe the gap.');
+      }
+    }
+  }
+
+  if (classification === 'UI-sensitive') {
+    if (!screenshotChecked) {
+      needsVisualEvidence = true;
+      evidenceGaps.push('The screenshot/clip checkbox is not checked for this UI-sensitive change.');
+    }
+    if (!hasVisualEvidence) {
+      needsVisualEvidence = true;
+      evidenceGaps.push('The Screenshots / clips section does not contain an actual attachment or link.');
+    }
   }
 
   // ── Comment ──────────────────────────────────────────────────────────────
@@ -62,22 +138,43 @@ module.exports = async ({ github, context, core }) => {
   });
   const existing = comments.find(c => (c.body ?? '').includes(MARKER));
 
-  if (problems.length === 0) {
+  if (descriptionProblems.length === 0 && evidenceGaps.length === 0) {
     if (existing) {
       await github.rest.issues.deleteComment({ owner, repo, comment_id: existing.id });
     }
   } else {
-    const commentBody = [
-      MARKER,
-      '⚠️ **PR description — action needed**',
-      '',
-      'The following required sections are missing or incomplete. Please update the PR description to address them:',
-      '',
-      problems.map(p => `- ${p}`).join('\n'),
+    const commentLines = [MARKER];
+    if (descriptionProblems.length > 0) {
+      commentLines.push(
+        '⚠️ **PR description — action needed**',
+        '',
+        'The following required sections are missing or incomplete. Please update the PR description to address them:',
+        '',
+        descriptionProblems.map(problem => `- ${problem}`).join('\n'),
+      );
+    } else {
+      commentLines.push(
+        '⚠️ **PR description is complete; validation evidence is still outstanding**',
+        '',
+        `Changed-file classification: **${classification}**.`,
+      );
+    }
+    if (evidenceGaps.length > 0) {
+      commentLines.push(
+        '',
+        '**Author-reported runtime / visual state**',
+        '',
+        evidenceGaps.map(gap => `- ${gap}`).join('\n'),
+        '',
+        'Checkboxes are author attestations. GitHub Actions results remain the execution evidence for CI; this check does not prove that a local command ran.',
+      );
+    }
+    commentLines.push(
       '',
       '---',
-      '_This comment is deleted automatically once all sections are complete._',
-    ].join('\n');
+      '_This comment updates automatically when the description or changed files change._',
+    );
+    const commentBody = commentLines.join('\n');
 
     if (existing) {
       await github.rest.issues.updateComment({ owner, repo, comment_id: existing.id, body: commentBody });
@@ -97,34 +194,47 @@ module.exports = async ({ github, context, core }) => {
       return true;
     } catch (e) {
       if (e.status === 404) return false;
+      if (e.status === 403) {
+        core.warning(`Could not inspect label "${name}" — token lacks label read access; skipping.`);
+        return false;
+      }
       throw e;
     }
   }
 
-  async function swapLabel(num, add, remove) {
-    if (await labelExists(add)) {
+  async function setLabel(name, wanted) {
+    if (wanted && await labelExists(name)) {
       try {
-        await github.rest.issues.addLabels({ owner, repo, issue_number: num, labels: [add] });
+        await github.rest.issues.addLabels({ owner, repo, issue_number: prNum, labels: [name] });
       } catch (e) {
         // Fail soft on a token that can't write labels so a label permission
         // problem never masks the actual description verdict.
-        if (e.status !== 403) throw e;
-        core.warning(`Could not add "${add}" — token lacks label write here; skipping.`);
+        if (e.status !== 403 && e.status !== 404) throw e;
+        core.warning(`Could not add "${name}" — label is unavailable or the token lacks label write access; skipping.`);
       }
+    } else if (wanted) {
+      core.warning(`Label "${name}" does not exist in the repo — skipping. Create it once to enable labelling.`);
     } else {
-      core.warning(`Label "${add}" does not exist in the repo — skipping. Create it once to enable labelling.`);
-    }
-    try {
-      await github.rest.issues.removeLabel({ owner, repo, issue_number: num, name: remove });
-    } catch (e) {
-      if (e.status !== 404 && e.status !== 410 && e.status !== 403) throw e;
+      try {
+        await github.rest.issues.removeLabel({ owner, repo, issue_number: prNum, name });
+      } catch (e) {
+        if (e.status !== 404 && e.status !== 410 && e.status !== 403) throw e;
+      }
     }
   }
 
-  if (problems.length === 0) {
-    await swapLabel(prNum, 'ready for review', 'needs work');
-  } else {
-    await swapLabel(prNum, 'needs work', 'ready for review');
-    core.setFailed(`PR description has ${problems.length} issue(s) — see bot comment for details.`);
+  const descriptionComplete = descriptionProblems.length === 0;
+  const evidenceComplete = evidenceGaps.length === 0;
+  const isDraft = Boolean(context.payload.pull_request.draft);
+  await setLabel(
+    'ready for review',
+    descriptionComplete && evidenceComplete && !isDraft,
+  );
+  await setLabel('needs work', !descriptionComplete);
+  await setLabel('needs runtime validation', needsRuntimeValidation);
+  await setLabel('needs visual evidence', needsVisualEvidence);
+
+  if (!descriptionComplete) {
+    core.setFailed(`PR description has ${descriptionProblems.length} issue(s) — see bot comment for details.`);
   }
 };

.github/workflows/pr-description-check.yml

@@ -5,7 +5,11 @@ on:
   # works on fork PRs. Safe here: the checkout pins to the base branch (no fork
   # code runs) and the scripts only read context.payload and call the GitHub API.
   pull_request_target:  # zizmor: ignore[dangerous-triggers]
-    types: [opened, edited, synchronize, reopened, ready_for_review]
+    types: [opened, edited, synchronize, reopened, ready_for_review, converted_to_draft]
+
+concurrency:
+  group: pr-description-${{ github.event.pull_request.number }}
+  cancel-in-progress: true
 
 # Default-deny at the workflow level; each job opts into only the scopes it needs.
 # Note: modifying a PR's labels/comments needs pull-requests:write even though the
@@ -59,12 +63,14 @@ jobs:
 
   check-mergeable:
     name: Flag unmergeable PRs
+    needs: check-description
     runs-on: ubuntu-latest
     permissions:
       pull-requests: write
       issues: write
-    # Skip bots: they open PRs programmatically and have their own process.
-    if: github.event.pull_request.user.type != 'Bot'
+    # Run after description validation failures, but never from an obsolete
+    # workflow run canceled by a newer PR event.
+    if: ${{ !cancelled() && github.event.pull_request.user.type != 'Bot' }}
     steps:
       - uses: actions/github-script@3a2844b7e9c422d3c10d287c895573f7108da1b3  # v9.0.0
         with:

ACKNOWLEDGMENTS.md

@@ -65,6 +65,16 @@ Vendored in `static/lib/` and served directly:
 | [jsPDF](https://github.com/parallax/jsPDF) (bundled in html2pdf) | PDF generation | MIT |
 | [html2canvas](https://github.com/niklasvh/html2canvas) (bundled in html2pdf) | DOM → canvas rasterization | MIT |
 | [node-qrcode](https://github.com/soldair/node-qrcode) (`qrcode.min.js`) | QR-code rendering (2FA setup) | MIT |
+| [KaTeX](https://github.com/KaTeX/KaTeX) v0.16.22 (`katex/katex.min.{js,css}` + `katex/fonts/*.woff2`) | Math typesetting | MIT ([`licenses/KaTeX-MIT-LICENSE.txt`](licenses/KaTeX-MIT-LICENSE.txt)) |
+| [Mermaid](https://github.com/mermaid-js/mermaid) v11.16.1 (`mermaid.min.js`) | Diagrams from text | MIT ([`licenses/Mermaid-MIT-LICENSE.txt`](licenses/Mermaid-MIT-LICENSE.txt)) |
+
+KaTeX and Mermaid are loaded on first use by `static/js/markdown.js` rather than
+from `index.html`, so a session that renders no math and no diagram never fetches
+either. Only the `.woff2` KaTeX fonts are shipped, matching `static/fonts/`; the
+`.woff` and `.ttf` variants its stylesheet also lists are never requested by a
+browser that supports `woff2`. The bundles are the published npm artifacts,
+unmodified — `.gitattributes` turns the whitespace check off for `static/lib/`
+so they can stay byte-identical to upstream.
 
 ## Front-end libraries loaded at runtime (CDN)
 
@@ -72,8 +82,6 @@ Referenced from `cdn.jsdelivr.net` / `cdnjs.cloudflare.com` at runtime — not v
 
 | Library | Purpose | License |
 |---|---|---|
-| [KaTeX](https://github.com/KaTeX/KaTeX) 0.16.22 | Math typesetting | MIT |
-| [Mermaid](https://github.com/mermaid-js/mermaid) 11 | Diagrams from text | MIT |
 | [Pyodide](https://github.com/pyodide/pyodide) 0.27.5 | In-browser Python runtime | MPL-2.0 |
 | [PDFObject](https://github.com/pipwerks/PDFObject) 2.1.1 | Inline PDF embedding | MIT |
 

README.md

@@ -59,15 +59,20 @@ Help is welcome. The best entry points are fresh-install testing, provider setup
 
 ## Security
 
-Odysseus is a self-hosted workspace with powerful local tools. Keep auth enabled, keep private data out of Git, and do not expose raw model/service ports publicly. Deployment details are in the [setup guide](docs/setup.md#security-notes).
+Odysseus is a self-hosted workspace with powerful local tools. Keep auth enabled, keep private data out of Git, and do not expose raw model/service ports publicly.
+
+- Keep `AUTH_ENABLED=true` for any network-accessible deployment.
+- Keep `LOCALHOST_BYPASS=false` outside local development.
+
+Deployment details are in the [setup guide](docs/setup.md#security-notes).
 
 ## Star History
 
-<a href="https://www.star-history.com/?repos=odysseus-dev%2Fodysseus&type=date&legend=top-left">
+<a href="https://star-history.dera.page/#odysseus-dev/odysseus&type=date&legend=top-left">
  <picture>
-   <source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/chart?repos=odysseus-dev/odysseus&type=date&theme=dark&legend=top-left" />
-   <source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/chart?repos=odysseus-dev/odysseus&type=date&legend=top-left" />
-   <img alt="Star History Chart" src="https://api.star-history.com/chart?repos=odysseus-dev/odysseus&type=date&legend=top-left" />
+   <source media="(prefers-color-scheme: dark)" srcset="https://star-history.dera.page/svg?repos=odysseus-dev/odysseus&type=date&theme=dark&legend=top-left" />
+   <source media="(prefers-color-scheme: light)" srcset="https://star-history.dera.page/svg?repos=odysseus-dev/odysseus&type=date&legend=top-left" />
+   <img alt="Star History Chart" src="https://star-history.dera.page/svg?repos=odysseus-dev/odysseus&type=date&legend=top-left" />
  </picture>
 </a>
 

SECURITY.md

@@ -10,7 +10,7 @@ Security fixes are handled on the default branch until formal releases are cut.
 
 - Keep `AUTH_ENABLED=true` for any network-accessible deployment.
 - Keep `LOCALHOST_BYPASS=false` outside local development.
-- Set `SECURE_COOKIES=true` when Odysseus is served through HTTPS by a trusted reverse proxy or private access gateway.
+- Leave `SECURE_COOKIES` unset unless you need to override it: session cookies are marked `Secure` whenever the request arrives over HTTPS. Set `SECURE_COOKIES=true` to force it on (for a proxy Odysseus cannot see the scheme of), or `SECURE_COOKIES=false` to force it off while you still serve plain HTTP alongside HTTPS.
 - Use HTTPS when exposing the app beyond localhost.
 - Put the authenticated Odysseus web/API entrypoint behind a trusted reverse proxy or private access layer such as Cloudflare Access, Tailscale, or a VPN.
 - Keep ChromaDB, SearXNG, ntfy, Ollama, vLLM, llama.cpp, databases, and raw model/provider APIs internal-only.

THREAT_MODEL.md

@@ -37,7 +37,7 @@ Non-admin defaults are in `core/auth.py:DEFAULT_PRIVILEGES`. Tool enforcement is
 
 - **Sessions:** bcrypt passwords, 7-day session tokens stored atomically in `data/sessions.json` via `core/atomic_io.py`.
 - **2FA:** TOTP with 8 single-use backup codes. Verified after password check, before session issuance.
-- **Reserved usernames:** `internal-tool`, `api`, `demo`, `system` cannot be registered or renamed into. Defined in `core/auth.py:RESERVED_USERNAMES`.
+- **Reserved usernames:** request sentinels and the Default/Local storage owner cannot be registered or renamed into. Defined in `core/auth.py:RESERVED_USERNAMES`.
   - `internal-tool` is security-critical: `core/middleware.py:require_admin` treats any request where `request.state.current_user == "internal-tool"` as the in-process tool loopback and grants admin unconditionally. A real account with that name would silently pass every `require_admin` check.
 - **Orphan sessions:** `validate_token` re-checks that the user record still exists on every call. A deleted user's cookie is dropped on next request rather than continuing to authenticate.
 

app.py

@@ -67,7 +67,13 @@ from core.constants import (
     REQUEST_TIMEOUT, OPENAI_API_KEY, AUTH_FILE,
 )
 from core.database import SessionLocal, ApiToken
-from core.middleware import SecurityHeadersMiddleware, is_cors_preflight
+from core.middleware import (
+    SecurityHeadersMiddleware,
+    get_application_route_path,
+    is_cors_preflight,
+    path_is_route_or_child,
+    with_asgi_root_path,
+)
 from core.auth import AuthManager, normalize_known_username
 from core.exceptions import (
     SessionNotFoundError, InvalidFileUploadError,
@@ -78,6 +84,7 @@ import bcrypt as _bcrypt
 
 from src.app_helpers import abs_join, serve_html_with_nonce
 from src.generated_images import GENERATED_IMAGE_HEADERS, resolve_generated_image_path
+from src.owner_identity import auth_disabled
 from starlette.responses import RedirectResponse
 
 # ========= LOGGING =========
@@ -248,7 +255,7 @@ from routes.auth_routes import setup_auth_routes, SESSION_COOKIE
 
 auth_manager = AuthManager()
 app.state.auth_manager = auth_manager
-AUTH_ENABLED = os.getenv("AUTH_ENABLED", "true").lower() != "false"
+AUTH_ENABLED = not auth_disabled()
 LOCALHOST_BYPASS = os.getenv("LOCALHOST_BYPASS", "false").lower() == "true"
 if LOCALHOST_BYPASS:
     logger.warning("LOCALHOST_BYPASS is enabled, loopback requests bypass authentication. Do not expose this instance to a network.")
@@ -284,7 +291,7 @@ if AUTH_ENABLED:
     def _is_auth_exempt(path: str) -> bool:
         if path in AUTH_EXEMPT_EXACT:
             return True
-        if any(path.startswith(p) for p in AUTH_EXEMPT_PREFIXES):
+        if any(path_is_route_or_child(path, p) for p in AUTH_EXEMPT_PREFIXES):
             return True
         return any(p.match(path) for p in AUTH_EXEMPT_PATTERNS)
 
@@ -355,7 +362,7 @@ if AUTH_ENABLED:
 
     class AuthMiddleware(BaseHTTPMiddleware):
         async def dispatch(self, request: Request, call_next):
-            path = request.url.path
+            path = get_application_route_path(request.scope)
             # A genuine CORS preflight (OPTIONS + Access-Control-Request-Method)
             # carries no credentials by design and must reach CORSMiddleware to be
             # answered. AuthMiddleware is the outermost middleware, so gating the
@@ -399,7 +406,10 @@ if AUTH_ENABLED:
             if not auth_manager.is_configured:
                 # No users yet — redirect to login for first-time setup
                 if not path.startswith("/api/"):
-                    return RedirectResponse(url="/login", status_code=302)
+                    return RedirectResponse(
+                        url=with_asgi_root_path(request.scope, "/login"),
+                        status_code=302,
+                    )
                 return JSONResponse(status_code=401, content={"error": "Setup required"})
 
             # --- Bearer token auth (API tokens for external integrations) ---
@@ -461,7 +471,10 @@ if AUTH_ENABLED:
             if not auth_manager.validate_token(token):
                 if path.startswith("/api/"):
                     return JSONResponse(status_code=401, content={"error": "Not authenticated"})
-                return RedirectResponse(url="/login", status_code=302)
+                return RedirectResponse(
+                    url=with_asgi_root_path(request.scope, "/login"),
+                    status_code=302,
+                )
 
             # Attach current username to request state for downstream routes
             request.state.current_user = auth_manager.get_username_for_token(token)
@@ -630,13 +643,24 @@ app.include_router(auth_router)
 
 @app.post("/api/activity/heartbeat")
 async def activity_heartbeat():
-    from src.interactive_gate import mark_browser_activity
+    from src.interactive_gate import (
+        mark_browser_activity,
+        maybe_stop_background_tasks_for_heartbeat,
+    )
+
     await mark_browser_activity()
+
     async def _stop_background():
         try:
-            await task_scheduler.stop_background_tasks_for_foreground(reason="browser heartbeat")
+            await maybe_stop_background_tasks_for_heartbeat(
+                task_scheduler.stop_background_tasks_for_foreground
+            )
         except Exception:
-            logging.getLogger("app.foreground_gate").debug("heartbeat task stop failed", exc_info=True)
+            logging.getLogger("app.foreground_gate").debug(
+                "heartbeat task stop failed",
+                exc_info=True,
+            )
+
     asyncio.create_task(_stop_background())
     return {"ok": True}
 
@@ -760,7 +784,7 @@ from src.task_scheduler import TaskScheduler
 task_scheduler = TaskScheduler(session_manager)
 from src.event_bus import set_task_scheduler
 set_task_scheduler(task_scheduler)
-from routes.task_routes import setup_task_routes
+from routes.task.task_routes import setup_task_routes
 app.include_router(setup_task_routes(task_scheduler))
 
 from routes.assistant_routes import setup_assistant_routes

build-macos-app.sh

@@ -73,6 +73,10 @@ cat > "$APP/Contents/MacOS/$APP_NAME.tmpl" <<'LAUNCHER'
 INSTALL_DIR="__INSTALL_DIR__"
 PORT="__PORT__"
 URL="http://127.0.0.1:${PORT}"
+# uvicorn is started with --port below, but APP_PORT is what the app itself
+# reads when it needs to build a URL for this instance (internal_api_base(),
+# companion pairing, the MCP OAuth callback), so export it as well.
+export APP_PORT="$PORT"
 export PATH="/opt/homebrew/bin:/usr/local/bin:/usr/bin:/bin:$PATH"
 
 UVICORN="$INSTALL_DIR/venv/bin/uvicorn"

companion/pairing.py

@@ -6,11 +6,14 @@ units so the route layer stays thin and the logic is directly testable.
 
 from __future__ import annotations
 
+import ipaddress
 import json
 import os
+import re
 import secrets
 import socket
 import uuid
+from urllib.parse import urlsplit
 
 import bcrypt
 
@@ -20,6 +23,102 @@ PAIRING_VERSION = 1
 COMPANION_SCOPE = "chat"
 
 
+_COMPANION_IPV4_NETWORKS = tuple(
+    ipaddress.ip_network(cidr)
+    for cidr in (
+        "10.0.0.0/8",
+        "100.64.0.0/10",
+        "127.0.0.0/8",
+        "169.254.0.0/16",
+        "172.16.0.0/12",
+        "192.168.0.0/16",
+    )
+)
+_DNS_LABEL_RE = re.compile(r"[a-z0-9](?:[a-z0-9-]{0,61}[a-z0-9])?\Z")
+
+
+def _valid_companion_client_host(host: str) -> bool:
+    """Match the host forms supported by the current v1 Expo client."""
+    if not host or len(host) > 253 or not host.isascii() or "%" in host:
+        return False
+
+    try:
+        address = ipaddress.ip_address(host)
+    except ValueError:
+        labels = host.split(".")
+        if any(not _DNS_LABEL_RE.fullmatch(label) for label in labels):
+            return False
+        if any(label.startswith("xn--") for label in labels):
+            return False
+        # WHATWG URL parsers treat a decimal or ``0x`` single-label hostname
+        # as an IPv4 number even though Python's strict ``ipaddress`` parser
+        # rejects that spelling.  The v1 client interpolates this host back
+        # into a URL, so accepting e.g. ``134744072`` would make the phone send
+        # its bearer token to public 8.8.8.8.  Keep DNS labels unambiguous.
+        if len(labels) == 1 and (
+            labels[0].isdigit()
+            or re.fullmatch(r"0x[0-9a-f]*", labels[0]) is not None
+        ):
+            return False
+        return len(labels) == 1 or (len(labels) >= 2 and labels[-1] == "local")
+
+    return isinstance(address, ipaddress.IPv4Address) and any(
+        address in network for network in _COMPANION_IPV4_NETWORKS
+    )
+
+
+def parse_companion_base_url(value: str) -> tuple[str, int]:
+    """Validate a v1 companion address and return its legacy (host, port).
+
+    The deployed client understands only HTTP plus a LAN-style host and port.
+    Reject anything outside that exact contract instead of advertising a URL
+    the client would reject, downgrade, or interpret differently.
+    """
+    if not isinstance(value, str) or not value:
+        raise ValueError("COMPANION_BASE_URL must be a canonical HTTP LAN origin")
+    if not value.isascii():
+        raise ValueError("COMPANION_BASE_URL must contain only ASCII characters")
+    if any(
+        ord(char) <= 32 or ord(char) == 127 or char in {"\\", "%"}
+        for char in value
+    ):
+        raise ValueError(
+            "COMPANION_BASE_URL contains a forbidden character"
+        )
+
+    try:
+        parsed = urlsplit(value)
+        port = parsed.port
+    except ValueError as exc:
+        raise ValueError("COMPANION_BASE_URL must be a valid HTTP LAN origin") from exc
+
+    host = parsed.hostname
+    if parsed.scheme.lower() != "http" or not parsed.netloc or not host:
+        raise ValueError("COMPANION_BASE_URL must be a canonical HTTP LAN origin")
+    if parsed.username is not None or parsed.password is not None:
+        raise ValueError("COMPANION_BASE_URL must not contain credentials")
+    if parsed.path or parsed.query or parsed.fragment:
+        raise ValueError("COMPANION_BASE_URL must not contain a path, query, or fragment")
+    if port is not None and not 1 <= port <= 65535:
+        raise ValueError("COMPANION_BASE_URL port must be between 1 and 65535")
+    if not _valid_companion_client_host(host):
+        raise ValueError("COMPANION_BASE_URL host is not supported by companion v1")
+
+    netloc = f"{host}:{port}" if port is not None else host
+    origin = f"http://{netloc}"
+    if value != origin:
+        raise ValueError("COMPANION_BASE_URL must be a canonical HTTP LAN origin")
+    return host, port or 80
+
+
+def configured_companion_origin() -> tuple[str, int] | None:
+    """Return the validated operator-configured v1 address, if any."""
+    value = os.environ.get("COMPANION_BASE_URL")
+    if value is None or value == "":
+        return None
+    return parse_companion_base_url(value)
+
+
 def default_port() -> int:
     """Best guess at the port the server is reachable on. Callers that know the
     real request port should pass it explicitly."""

companion/routes.py

@@ -23,7 +23,7 @@ from fastapi import APIRouter, HTTPException, Request
 from fastapi.responses import HTMLResponse
 
 from core.middleware import require_admin
-from src.auth_helpers import get_current_user
+from src.auth_helpers import _auth_disabled, get_current_user
 
 from companion import pairing as _pairing
 
@@ -113,8 +113,9 @@ def setup_companion_routes() -> APIRouter:
         The stock /api/models route scopes to get_current_user, which for a
         bearer token is the sandboxed pseudo-user "api" (owns nothing). Here we
         scope to the token's real owner instead, plus legacy null-owner shared
-        rows -- the same rule as owner_filter. Read-only; never returns api_key
-        material.
+        rows -- the same rule as owner_filter. Explicit auth-disabled mode keeps
+        the stock route's single-user all-endpoints view. Read-only; never
+        returns api_key material.
         """
         require_models_scope(request)
         import json as _json
@@ -123,6 +124,11 @@ def setup_companion_routes() -> APIRouter:
         from src.endpoint_resolver import build_chat_url
 
         owner = token_owner(request)
+        single_user_mode = (
+            owner is None
+            and not getattr(request.state, "api_token", False)
+            and _auth_disabled()
+        )
         out = []
         db = SessionLocal()
         try:
@@ -133,7 +139,7 @@ def setup_companion_routes() -> APIRouter:
             if owner:
                 q = q.filter((ModelEndpoint.owner == owner) | (ModelEndpoint.owner == None))  # noqa: E711
             for ep in q.all():
-                if not owner_can_see(ep.owner, owner):
+                if not single_user_mode and not owner_can_see(ep.owner, owner):
                     continue
                 try:
                     model_ids = _json.loads(ep.cached_models) if ep.cached_models else []
@@ -194,19 +200,27 @@ def setup_companion_routes() -> APIRouter:
         the code works immediately, no restart. `?format=json` returns the
         payload for an in-app pairing screen."""
         require_admin(request)
+        try:
+            configured_origin = _pairing.configured_companion_origin()
+        except ValueError as exc:
+            raise HTTPException(500, str(exc)) from None
         owner = get_current_user(request)
         invalidate = getattr(request.app.state, "invalidate_token_cache", None)
         token_id, raw_token = mint_pairing_token(owner, invalidate)
 
-        hosts = _pairing.lan_ip_candidates()
-        host = hosts[0] if hosts else "127.0.0.1"
-        port = request.url.port or _pairing.default_port()
+        if configured_origin:
+            host, port = configured_origin
+            hosts = [host]
+        else:
+            hosts = _pairing.lan_ip_candidates()
+            host = hosts[0] if hosts else "127.0.0.1"
+            port = request.url.port or _pairing.default_port()
         payload = _pairing.pairing_payload(host, port, raw_token)
         qr = _pairing.pairing_qr_png_data_uri(payload)
         qr_ok = bool(qr and qr.startswith("data:image/png;base64,"))
 
         if (request.query_params.get("format") or "").lower() == "json":
-            return {
+            response = {
                 "host": host,
                 "port": port,
                 "token": raw_token,
@@ -215,6 +229,7 @@ def setup_companion_routes() -> APIRouter:
                 "payload": payload,
                 "qr": qr if qr_ok else None,
             }
+            return response
 
         import json as _json
         payload_json = _json.dumps(payload, separators=(",", ":"))

core/auth.py

@@ -20,7 +20,6 @@ logger = logging.getLogger(__name__)
 
 
 from core.atomic_io import atomic_write_json as _atomic_write_json  # noqa: E402
-from core.middleware import INTERNAL_TOOL_USER  # noqa: E402
 
 DEFAULT_PRIVILEGES = {
     "can_use_agent": True,
@@ -49,24 +48,18 @@ ADMIN_PRIVILEGES["allowed_models_restricted"] = False
 ADMIN_PRIVILEGES["block_all_models"] = False
 
 from src.constants import AUTH_FILE, PASSWORD_MIN_LENGTH
+from src.owner_identity import RESERVED_AUTH_USERNAMES
 DEFAULT_AUTH_PATH = AUTH_FILE
 TOKEN_TTL = 60 * 60 * 24 * 7  # 7 days
 
-# Usernames the auth + middleware layer reserve as internal "synthetic owner"
-# sentinels; they must never belong to a real account. The most dangerous is
-# "internal-tool": `core.middleware.require_admin` treats any request whose
-# `current_user == "internal-tool"` as the in-process tool loopback and grants
-# admin, and because the cookie auth path sets `current_user` to the raw
-# username, an account literally named "internal-tool" would be silently
-# treated as an admin by every `require_admin`-gated route. "api" collides with
-# the bearer-token owner-attribution sentinel. "demo"/"system" round out the
-# synthetic-owner set the rest of the codebase already special-cases (see
-# `_SYNTHETIC_OWNERS` in routes/assistant_routes.py and the matching guards in
-# src/task_scheduler.py / routes/research_routes.py) — a real account with one
-# of those names would be denied an assistant and inconsistently owner-scoped.
-# Refuse to create or rename into any of them so the sentinels can't be
-# impersonated. (Keep this in sync with that synthetic-owner set.)
-RESERVED_USERNAMES = frozenset({INTERNAL_TOOL_USER, "api", "demo", "system"})
+# Usernames the auth + middleware layer reserves for request sentinels and
+# internal storage owners; they must never belong to a real login account.
+# "internal-tool" is the most dangerous because `core.middleware.require_admin`
+# treats it as the in-process tool loopback. "api" collides with bearer-token
+# attribution. "demo"/"system" are synthetic owners already special-cased by
+# scheduler/assistant/research paths. The Default/Local owner is a storage
+# bucket for explicit auth-disabled no-login mode, not a login username.
+RESERVED_USERNAMES = frozenset(RESERVED_AUTH_USERNAMES)
 
 
 def normalize_known_username(users: Dict[str, Any], username: str | None) -> Optional[str]:

core/database.py

@@ -1491,8 +1491,25 @@ def _migrate_assign_legacy_owner():
             with open(prefs_path, "r", encoding="utf-8") as f:
                 prefs = _json.load(f)
             if "_users" not in prefs and prefs:
-                # Flat format → nest under admin user
-                new_prefs = {"_users": {admin_user: prefs}}
+                # Flat format → nest ordinary preferences under the admin
+                # user. Foreground fallback is an explicit per-owner opt-in,
+                # so auth-disabled consent must remain inert at the flat root
+                # rather than becoming consent for the first named owner.
+                foreground_keys = {
+                    "foreground_fallback_enabled",
+                    "foreground_model_fallbacks",
+                }
+                named_prefs = {
+                    key: value
+                    for key, value in prefs.items()
+                    if key not in foreground_keys
+                }
+                new_prefs = {
+                    key: prefs[key]
+                    for key in foreground_keys
+                    if key in prefs
+                }
+                new_prefs["_users"] = {admin_user: named_prefs}
                 with open(prefs_path, "w", encoding="utf-8") as f:
                     _json.dump(new_prefs, f, indent=2)
                 logger.info(f"Migrated user_prefs.json to per-user format under '{admin_user}'")

core/middleware.py

@@ -3,10 +3,14 @@
 
 import os
 import secrets
+from collections.abc import Mapping
 
 from fastapi import HTTPException, Request
 from starlette.middleware.base import BaseHTTPMiddleware
 from starlette.responses import Response
+from starlette.routing import get_route_path
+
+from src.owner_identity import INTERNAL_TOOL_USER, auth_disabled
 
 
 # Per-process token that lets the in-app tool layer hit admin-gated
@@ -15,8 +19,30 @@ from starlette.responses import Response
 # same value from this module. Never persisted or exposed externally.
 INTERNAL_TOOL_TOKEN = os.environ.get("ODYSSEUS_INTERNAL_TOKEN") or secrets.token_hex(32)
 INTERNAL_TOOL_HEADER = "X-Odysseus-Internal-Token"
-# Pseudo-username on in-process tool-loopback requests; require_admin trusts it and it is reserved.
-INTERNAL_TOOL_USER = "internal-tool"
+
+
+def get_application_route_path(scope: Mapping[str, object]) -> str:
+    """Return the application-relative path used by Starlette routing.
+
+    Uvicorn prefixes ``scope["path"]`` with a configured ASGI ``root_path``;
+    Starlette removes that prefix before matching routes. Middleware policy
+    must use the same path form or a deployment prefix can change which policy
+    applies to an otherwise unchanged application route.
+    """
+    return get_route_path(scope)
+
+
+def with_asgi_root_path(scope: Mapping[str, object], path: str) -> str:
+    """Prefix an application path for a client-facing redirect target."""
+    root_path = scope.get("root_path", "")
+    if not isinstance(root_path, str) or not root_path:
+        return path
+    return f"{root_path.rstrip('/')}{path}"
+
+
+def path_is_route_or_child(path: str, prefix: str) -> bool:
+    """Return whether ``path`` is exactly ``prefix`` or below that route."""
+    return path == prefix or path.startswith(prefix + "/")
 
 
 def is_cors_preflight(method: str, headers) -> bool:
@@ -47,7 +73,7 @@ def require_admin(request: Request):
         pass
 
     auth_mgr = getattr(request.app.state, "auth_manager", None)
-    if os.getenv("AUTH_ENABLED", "true").lower() == "false":
+    if auth_disabled():
         return
     if not auth_mgr or not auth_mgr.is_configured:
         raise HTTPException(403, "Admin only")

core/session_manager.py

@@ -14,6 +14,8 @@ import logging
 from datetime import datetime, timezone, timedelta
 from typing import Dict, Optional
 
+from sqlalchemy import func
+
 from .database import Session as DbSession, ChatMessage as DbChatMessage, Document as DbDocument, SessionLocal, utcnow_naive
 from .models import Session, ChatMessage
 from src.attachment_refs import persistable_message_content
@@ -92,14 +94,28 @@ class SessionManager:
         try:
             db_sessions = db.query(DbSession).filter(
                 DbSession.archived == False,
-                DbSession.message_count > 0,
+                DbSession.messages.any(),
             ).order_by(DbSession.last_accessed.desc()).limit(100).all()
 
+            # message_count is derived metadata and can drift after interrupted
+            # or legacy writes. Count only the bounded discovery set so startup
+            # remains metadata-only while lazy hydration sees an authoritative
+            # positive count for every discovered non-empty session.
+            message_counts = {}
+            if db_sessions:
+                message_counts = dict(
+                    db.query(DbChatMessage.session_id, func.count(DbChatMessage.id))
+                    .filter(DbChatMessage.session_id.in_([row.id for row in db_sessions]))
+                    .group_by(DbChatMessage.session_id)
+                    .all()
+                )
+
             loaded_count = 0
             for db_session in db_sessions:
                 try:
                     session = self._db_to_session_meta(db_session)
                     if session is not None:
+                        session.message_count = message_counts[db_session.id]
                         self.sessions[db_session.id] = session
                         loaded_count += 1
                 except Exception as e:

docker-compose.gpu-amd.yml

@@ -46,10 +46,11 @@ services:
       - DATABASE_URL=${DATABASE_URL:-sqlite:///./data/app.db}
       - AUTH_ENABLED=${AUTH_ENABLED:-true}
       - LOCALHOST_BYPASS=${LOCALHOST_BYPASS:-false}
+      - COMPANION_BASE_URL=${COMPANION_BASE_URL:-}
       - ODYSSEUS_ADMIN_USER=${ODYSSEUS_ADMIN_USER:-admin}
       - ODYSSEUS_ADMIN_PASSWORD=${ODYSSEUS_ADMIN_PASSWORD:-}
       - ALLOWED_ORIGINS=${ALLOWED_ORIGINS:-http://localhost,http://127.0.0.1}
-      - SECURE_COOKIES=${SECURE_COOKIES:-false}
+      - SECURE_COOKIES=${SECURE_COOKIES:-}
       - EMBEDDING_URL=${EMBEDDING_URL:-}
       - EMBEDDING_MODEL=${EMBEDDING_MODEL:-}
       - EMBEDDING_API_KEY=${EMBEDDING_API_KEY:-}
@@ -74,6 +75,11 @@ services:
       - GOOGLE_OAUTH_CLIENT_ID=${GOOGLE_OAUTH_CLIENT_ID:-}
       - GOOGLE_OAUTH_CLIENT_SECRET=${GOOGLE_OAUTH_CLIENT_SECRET:-}
       - GOOGLE_OAUTH_REDIRECT_URI=${GOOGLE_OAUTH_REDIRECT_URI:-}
+      # Externally reachable origin for MCP OAuth callbacks. The container
+      # always listens on 7000 and cannot see the host port map above, so
+      # remote MCP OAuth needs this set whenever the browser reaches
+      # Odysseus on anything other than http://localhost:7000.
+      - OAUTH_REDIRECT_BASE_URL=${OAUTH_REDIRECT_BASE_URL:-}
       - TAVILY_API_KEY=${TAVILY_API_KEY:-}
       - SERPER_API_KEY=${SERPER_API_KEY:-}
       # PUID / PGID — the user/group the container drops to before
@@ -129,12 +135,17 @@ services:
           fi
           sed "s|__SEARXNG_SECRET__|$$secret|g" /tmp/searxng-settings.yml.template > /etc/searxng/settings.yml
         fi
+        # Advisory: a settings file the migration cannot parse or rewrite must
+        # not be what stops searxng from booting. It explains itself on stderr
+        # and we carry on, letting searxng report anything genuinely wrong.
+        /usr/local/searxng/.venv/bin/python /tmp/migrate-searxng-settings.py /etc/searxng/settings.yml || true
         exec /usr/local/searxng/entrypoint.sh
     ports:
       - "127.0.0.1:8080:8080"
     volumes:
       - searxng-data:/etc/searxng
       - ./config/searxng/settings.yml:/tmp/searxng-settings.yml.template:ro,z
+      - ./scripts/migrate_searxng_settings.py:/tmp/migrate-searxng-settings.py:ro,z
     environment:
       - SEARXNG_BASE_URL=http://localhost:8080/
       - SEARXNG_SECRET=${SEARXNG_SECRET:-}

docker-compose.gpu-nvidia.yml

@@ -45,10 +45,11 @@ services:
       - DATABASE_URL=${DATABASE_URL:-sqlite:///./data/app.db}
       - AUTH_ENABLED=${AUTH_ENABLED:-true}
       - LOCALHOST_BYPASS=${LOCALHOST_BYPASS:-false}
+      - COMPANION_BASE_URL=${COMPANION_BASE_URL:-}
       - ODYSSEUS_ADMIN_USER=${ODYSSEUS_ADMIN_USER:-admin}
       - ODYSSEUS_ADMIN_PASSWORD=${ODYSSEUS_ADMIN_PASSWORD:-}
       - ALLOWED_ORIGINS=${ALLOWED_ORIGINS:-http://localhost,http://127.0.0.1}
-      - SECURE_COOKIES=${SECURE_COOKIES:-false}
+      - SECURE_COOKIES=${SECURE_COOKIES:-}
       - EMBEDDING_URL=${EMBEDDING_URL:-}
       - EMBEDDING_MODEL=${EMBEDDING_MODEL:-}
       - EMBEDDING_API_KEY=${EMBEDDING_API_KEY:-}
@@ -73,6 +74,11 @@ services:
       - GOOGLE_OAUTH_CLIENT_ID=${GOOGLE_OAUTH_CLIENT_ID:-}
       - GOOGLE_OAUTH_CLIENT_SECRET=${GOOGLE_OAUTH_CLIENT_SECRET:-}
       - GOOGLE_OAUTH_REDIRECT_URI=${GOOGLE_OAUTH_REDIRECT_URI:-}
+      # Externally reachable origin for MCP OAuth callbacks. The container
+      # always listens on 7000 and cannot see the host port map above, so
+      # remote MCP OAuth needs this set whenever the browser reaches
+      # Odysseus on anything other than http://localhost:7000.
+      - OAUTH_REDIRECT_BASE_URL=${OAUTH_REDIRECT_BASE_URL:-}
       - TAVILY_API_KEY=${TAVILY_API_KEY:-}
       - SERPER_API_KEY=${SERPER_API_KEY:-}
       # PUID / PGID — the user/group the container drops to before
@@ -132,12 +138,17 @@ services:
           fi
           sed "s|__SEARXNG_SECRET__|$$secret|g" /tmp/searxng-settings.yml.template > /etc/searxng/settings.yml
         fi
+        # Advisory: a settings file the migration cannot parse or rewrite must
+        # not be what stops searxng from booting. It explains itself on stderr
+        # and we carry on, letting searxng report anything genuinely wrong.
+        /usr/local/searxng/.venv/bin/python /tmp/migrate-searxng-settings.py /etc/searxng/settings.yml || true
         exec /usr/local/searxng/entrypoint.sh
     ports:
       - "127.0.0.1:8080:8080"
     volumes:
       - searxng-data:/etc/searxng
       - ./config/searxng/settings.yml:/tmp/searxng-settings.yml.template:ro,z
+      - ./scripts/migrate_searxng_settings.py:/tmp/migrate-searxng-settings.py:ro,z
     environment:
       - SEARXNG_BASE_URL=http://localhost:8080/
       - SEARXNG_SECRET=${SEARXNG_SECRET:-}

docker-compose.yml

@@ -34,10 +34,11 @@ services:
       - DATABASE_URL=${DATABASE_URL:-sqlite:///./data/app.db}
       - AUTH_ENABLED=${AUTH_ENABLED:-true}
       - LOCALHOST_BYPASS=${LOCALHOST_BYPASS:-false}
+      - COMPANION_BASE_URL=${COMPANION_BASE_URL:-}
       - ODYSSEUS_ADMIN_USER=${ODYSSEUS_ADMIN_USER:-admin}
       - ODYSSEUS_ADMIN_PASSWORD=${ODYSSEUS_ADMIN_PASSWORD:-}
       - ALLOWED_ORIGINS=${ALLOWED_ORIGINS:-http://localhost,http://127.0.0.1}
-      - SECURE_COOKIES=${SECURE_COOKIES:-false}
+      - SECURE_COOKIES=${SECURE_COOKIES:-}
       - EMBEDDING_URL=${EMBEDDING_URL:-}
       - EMBEDDING_MODEL=${EMBEDDING_MODEL:-}
       - EMBEDDING_API_KEY=${EMBEDDING_API_KEY:-}
@@ -62,6 +63,11 @@ services:
       - GOOGLE_OAUTH_CLIENT_ID=${GOOGLE_OAUTH_CLIENT_ID:-}
       - GOOGLE_OAUTH_CLIENT_SECRET=${GOOGLE_OAUTH_CLIENT_SECRET:-}
       - GOOGLE_OAUTH_REDIRECT_URI=${GOOGLE_OAUTH_REDIRECT_URI:-}
+      # Externally reachable origin for MCP OAuth callbacks. The container
+      # always listens on 7000 and cannot see the host port map above, so
+      # remote MCP OAuth needs this set whenever the browser reaches
+      # Odysseus on anything other than http://localhost:7000.
+      - OAUTH_REDIRECT_BASE_URL=${OAUTH_REDIRECT_BASE_URL:-}
       - TAVILY_API_KEY=${TAVILY_API_KEY:-}
       - SERPER_API_KEY=${SERPER_API_KEY:-}
       # PUID / PGID — the user/group the container drops to before
@@ -110,12 +116,17 @@ services:
           fi
           sed "s|__SEARXNG_SECRET__|$$secret|g" /tmp/searxng-settings.yml.template > /etc/searxng/settings.yml
         fi
+        # Advisory: a settings file the migration cannot parse or rewrite must
+        # not be what stops searxng from booting. It explains itself on stderr
+        # and we carry on, letting searxng report anything genuinely wrong.
+        /usr/local/searxng/.venv/bin/python /tmp/migrate-searxng-settings.py /etc/searxng/settings.yml || true
         exec /usr/local/searxng/entrypoint.sh
     ports:
       - "127.0.0.1:8080:8080"
     volumes:
       - searxng-data:/etc/searxng
       - ./config/searxng/settings.yml:/tmp/searxng-settings.yml.template:ro,z
+      - ./scripts/migrate_searxng_settings.py:/tmp/migrate-searxng-settings.py:ro,z
     environment:
       - SEARXNG_BASE_URL=http://localhost:8080/
       - SEARXNG_SECRET=${SEARXNG_SECRET:-}

docs/setup.md

@@ -441,10 +441,19 @@ A grab-bag of small gotchas that otherwise turn into long debugging sessions.
 | Package | Feature unlocked |
 |---------|-----------------|
 | `faster-whisper` | Local speech-to-text (microphone -> text) via the "local" STT provider. |
+| `kokoro`, `soundfile` | Local Kokoro-82M text-to-speech on a CUDA GPU. The pinned Kokoro release supports Odysseus installs on Python 3.11-3.12; these packages are intentionally skipped on Python 3.13+ (including the Python 3.14 container image). |
 | `ddgs` | DuckDuckGo as a search provider option. |
 | `PyMuPDF` | PDF page rendering in the side viewer panel and form-filling. (Note: AGPL-3.0) |
 | `markitdown` | Office/EPUB document text extraction (converts .docx/.xlsx/.pptx/.xls/.epub to Markdown). |
 
+Install the optional set only when you need these features:
+
+```bash
+pip install -r requirements-optional.txt
+```
+
+The default Docker image currently uses Python 3.14, while Kokoro 0.9.4 declares Python `>=3.10,<3.13`. Odysseus itself continues to support Python 3.11+, but this pinned optional local-TTS feature requires a native Python 3.11 or 3.12 environment. Kokoro declares `torch`, but the local provider only activates when that torch build has CUDA and a GPU is visible; install the CUDA build appropriate for your host. Browser and configured endpoint TTS remain available on Python 3.13+ and in the container image.
+
 ### Faster, reproducible installs with uv (optional)
 [uv](https://docs.astral.sh/uv/) works as a drop-in replacement for the
 venv + pip steps in the native install guides, no project changes are needed but this change results in faster installs along with a lockfile for reproducible environments. After [installing `uv`](https://docs.astral.sh/uv/getting-started/installation/), use:
@@ -475,7 +484,7 @@ Odysseus is a self-hosted workspace with powerful local tools: shell access, fil
 
 - Keep `AUTH_ENABLED=true` for any network-accessible deployment.
 - Keep `LOCALHOST_BYPASS=false` outside local development.
-- Use `SECURE_COOKIES=true` when Odysseus is served through HTTPS by a trusted reverse proxy or private access gateway.
+- Leave `SECURE_COOKIES` unset unless you need to override it: session cookies are marked `Secure` whenever the request arrives over HTTPS. Use `SECURE_COOKIES=true` to force it on for a proxy whose scheme Odysseus cannot see, or `SECURE_COOKIES=false` to force it off while you still serve plain HTTP alongside HTTPS.
 - Do not expose it directly to the public internet without HTTPS and a trusted reverse proxy or private access layer.
 - Keep `.env`, `data/`, `logs/`, databases, uploads, generated media, backups, auth/session files, API keys, and model/provider tokens out of Git and private shares. They are ignored by default.
 - Review `data/auth.json` after first boot: disable open signup unless you intentionally want it, make only your own account admin, and keep demo/test accounts non-admin.
@@ -486,6 +495,14 @@ Odysseus is a self-hosted workspace with powerful local tools: shell access, fil
 - Keep ChromaDB, SearXNG, ntfy, Ollama, vLLM, llama.cpp, databases, and raw model/provider APIs internal-only. Expose only the authenticated Odysseus web/API entrypoint through your trusted proxy or private access layer.
 - Before publishing a fork, run `git status --short` and confirm no private files from `.env`, `data/`, `logs/`, uploads, backups, or local databases are staged.
 
+> **Upgrading an existing install:** `SECURE_COOKIES` used to default to
+> `false`, so an install set up before scheme derivation may still carry
+> `SECURE_COOKIES=false` in its own `.env`. That explicit value stays
+> authoritative, so HTTPS logins keep getting a non-`Secure` session cookie.
+> Pulling this change updates the tracked Compose files, but nothing rewrites
+> your `.env` — drop the line from it unless you deliberately serve plain HTTP
+> alongside HTTPS and want the escape hatch.
+
 ### Private or proxied deployments
 Odysseus serves plain HTTP on its app port. Docker Compose binds Odysseus and the bundled services to `127.0.0.1` by default, so a typical production/private setup is:
 
@@ -494,9 +511,162 @@ Odysseus serves plain HTTP on its app port. Docker Compose binds Odysseus and th
 3. Put the authenticated Odysseus web/API entrypoint behind that layer.
 4. Keep raw service and model ports internal-only.
 
-Cloudflare Access, Tailscale, Caddy, nginx, and Traefik can all fit this pattern; none are required by Odysseus. If your access layer reaches Odysseus on the same host, proxy to `http://127.0.0.1:7000` and keep `AUTH_ENABLED=true`, `LOCALHOST_BYPASS=false`, and `SECURE_COOKIES=true`.
+Cloudflare Access, Tailscale, Caddy, nginx, and Traefik can all fit this pattern; none are required by Odysseus. If your access layer reaches Odysseus on the same host, proxy to `http://127.0.0.1:7000` and keep `AUTH_ENABLED=true` and `LOCALHOST_BYPASS=false`. Any proxy that forwards `X-Forwarded-Proto: https` gets `Secure` session cookies without configuration, so `SECURE_COOKIES` only needs setting when you want to override that — force it on for a proxy that forwards no scheme at all, or off while you still serve plain HTTP.
 `ALLOWED_ORIGINS` lists exact permitted origins for cross-origin browser/API clients; ordinary same-origin reverse-proxy access usually does not need a special CORS entry.
 
+#### Faster over the network: HTTP/2
+
+The frontend is raw ES modules with no bundler, so a page load is a few hundred
+small same-origin requests. Over HTTP/1.1 browsers typically allow only a small
+number of concurrent connections per host (commonly around six), so many of
+those requests are serialized across multiple round trips. On localhost that
+costs almost nothing. Over a LAN, VPN, or remote link it can become a major
+part of load time, especially as latency increases.
+
+HTTP/2 multiplexes them onto one connection and the serialisation disappears.
+Odysseus needs no changes for this — uvicorn keeps speaking HTTP/1.1 on
+loopback and the proxy speaks HTTP/2 to the browser. Mainstream browsers
+negotiate HTTP/2 for normal web pages over TLS; they do not use the cleartext
+h2c mode here, so browser-facing HTTP/2 requires a certificate. The
+`--ssl-certfile` route in *HTTPS + LAN/Tailscale exposure* above gives you
+HTTPS but not HTTP/2 — uvicorn does not speak it.
+
+**1. Install Caddy.** See the [install docs](https://caddyserver.com/docs/install)
+for your platform; on macOS, `brew install caddy`.
+
+**2. Write a `Caddyfile`.** Pick the block that matches how you reach the
+machine. Replace `7000` if Odysseus listens elsewhere — the macOS start script
+uses `7860`.
+
+Public domain, Caddy obtains and renews the certificate itself:
+
+```
+odysseus.example.com {
+	reverse_proxy 127.0.0.1:7000
+}
+```
+
+Tailscale, no public DNS needed — `tailscale cert` issues a browser-trusted
+certificate for a tailnet name and writes `<domain>.crt` and `<domain>.key`:
+
+```bash
+tailscale cert myhost.tailnet-name.ts.net
+```
+
+```
+myhost.tailnet-name.ts.net {
+	tls /path/to/myhost.tailnet-name.ts.net.crt /path/to/myhost.tailnet-name.ts.net.key
+	reverse_proxy 127.0.0.1:7000
+}
+```
+
+LAN with your own certificate — same shape, your own files:
+
+```
+odysseus.lan {
+	tls /path/to/cert.pem /path/to/key.pem
+	reverse_proxy 127.0.0.1:7000
+}
+```
+
+Give `tls` absolute paths: a service starts in a working directory you did not
+choose. If port 443 is already taken, append a port to the site address
+(`odysseus.example.com:8443`) and use it in the URL. That alone does not free
+port 80 — Caddy still binds it for the HTTP-to-HTTPS redirect, and fails to
+start with `listen tcp :80: bind: address already in use` if something else
+holds it. Turn the redirect off with a global block at the top of the file:
+
+```
+{
+	auto_https disable_redirects
+}
+```
+
+**3. Run it in the foreground first:**
+
+```bash
+caddy run --config ./Caddyfile
+```
+
+Once that works, run it as a service:
+
+```bash
+brew services start caddy          # macOS — reads $(brew --prefix)/etc/Caddyfile, not ./Caddyfile
+sudo systemctl enable --now caddy  # Linux, if your package installed the unit
+```
+
+Odysseus's own service is unchanged; the proxy runs alongside it. Under Docker,
+run the proxy as another container, or on the host pointing at the published
+port.
+
+**4. Point Odysseus at the new origin** in `.env`, then restart it.
+
+A proxy that exposes the HTTPS request scheme to Odysseus needs no `SECURE_COOKIES` setting. Only force it on when the proxy cannot expose that scheme:
+
+```bash
+# only if the proxy cannot expose the external HTTPS scheme to Odysseus:
+SECURE_COOKIES=true
+# only if you use remote MCP servers with OAuth:
+OAUTH_REDIRECT_BASE_URL=https://odysseus.example.com
+```
+
+Gmail OAuth needs nothing here when the proxy runs on the same host: the
+redirect URI is built from the incoming request, and uvicorn rewrites the
+scheme from `X-Forwarded-Proto` for proxies it trusts — by default only
+`127.0.0.1`. A proxy in a separate container or on another machine is not
+trusted, so pin the URI there:
+
+```bash
+GOOGLE_OAUTH_REDIRECT_URI=https://odysseus.example.com/api/email/oauth/google/callback
+```
+
+(uvicorn's own `FORWARDED_ALLOW_IPS` widens that trust, but it has to be in the
+environment uvicorn starts with — `.env` is read by the app afterwards, too
+late for it to take effect.)
+
+**5. Confirm HTTP/2 is really on:**
+
+```bash
+curl -s -o /dev/null -w '%{http_version}\n' https://odysseus.example.com/
+# 2
+```
+
+The status code is not the thing to check here — a logged-out request redirects
+to the login page, so `curl -I` shows `HTTP/2 302`, and the `HTTP/2` prefix is
+the part that matters. The browser reports the same in the Network panel's
+Protocol column (`h2`); in Chrome and Firefox that column is hidden until you
+enable it by right-clicking the column headers.
+
+Three things bite when moving an existing install behind TLS:
+
+- Leave `SECURE_COOKIES` unset when Odysseus can see the external HTTPS scheme;
+  the cookie then follows the request automatically. If your proxy cannot expose
+  that scheme, set `SECURE_COOKIES=true` **at the same time** you stop serving
+  plain HTTP, not before. An explicit `true` applies to every login, so while an
+  HTTP entrypoint is still reachable the browser will reject the `Secure` cookie
+  there and login will appear to loop.
+- `OAUTH_REDIRECT_BASE_URL` defaults to `http://localhost:7000`. Unlike the
+  Gmail redirect URI it cannot be derived from a request — it is registered
+  with each MCP authorization server up front — so set it to the external
+  origin if you use remote MCP servers over OAuth.
+- Odysseus sends `Strict-Transport-Security` once it sees `X-Forwarded-Proto:
+  https`. HSTS applies to the whole hostname and ignores the port, so any other
+  plain-HTTP service on that same hostname becomes unreachable in browsers that
+  have visited Odysseus. Give Odysseus its own hostname, or strip the header at
+  the proxy (`header_down -Strict-Transport-Security` in Caddy).
+
+Server-sent events are not buffered by this configuration, so chat streaming
+arrives token by token; add `flush_interval -1` inside the `reverse_proxy`
+block if you want that pinned explicitly. nginx needs `proxy_buffering off;`
+for the same reason.
+
+Changing the external origin also affects state scoped to it. Service workers
+and their caches are origin-scoped, so moving to a different origin starts with
+a cold load. Cookies follow their own domain/path/security rules rather than
+being port-scoped: changing the hostname normally requires a new login, while
+changing only the scheme or port does not by itself guarantee that existing
+cookies disappear.
+
 Common internal-only ports from the default docs/compose setup:
 
 | Port | Service |
@@ -527,7 +697,7 @@ Key settings:
 | `AUTH_ENABLED` | `true` | Enable/disable login |
 | `LOCALHOST_BYPASS` | `false` | Development-only auth bypass for loopback requests. Keep false for shared/network deployments. |
 | `ALLOWED_ORIGINS` | `http://localhost,http://127.0.0.1` | Comma-separated exact permitted origins for cross-origin browser/API clients. |
-| `SECURE_COOKIES` | `false` | Set true when serving Odysseus through HTTPS at a trusted proxy or private access gateway. |
+| `SECURE_COOKIES` | derived from the request scheme | Marks session cookies `Secure` on HTTPS requests. Set true to force it on, false to force it off. |
 | `DATABASE_URL` | `sqlite:///./data/app.db` | Database connection string |
 | `CHROMADB_HOST` | `localhost` | ChromaDB host for vector memory. Docker overrides this to `chromadb`. |
 | `CHROMADB_PORT` | `8100` | ChromaDB port for manual host runs. Docker overrides this to `8000`. |

launch-windows.ps1

@@ -163,6 +163,10 @@ if (Test-Path $cudaBase) {
 }
 
 # 7. Start the server (use `python -m uvicorn` - bare `uvicorn` may not be on PATH)
+# -Port only reaches uvicorn as a flag. Everything that builds a URL for this
+# instance - internal_api_base(), companion pairing, the MCP OAuth callback -
+# reads APP_PORT, so set it too or they all assume 7000.
+$env:APP_PORT = $Port
 Write-Step ("Starting Odysseus at http://{0}:{1}" -f $BindHost, $Port)
 Write-Host "Press Ctrl+C to stop."
 Write-Host ""

licenses/KaTeX-MIT-LICENSE.txt

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2013-2020 Khan Academy and other contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

licenses/Mermaid-MIT-LICENSE.txt

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2014 - 2022 Knut Sveidqvist
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

mcp_servers/email_server.py

@@ -1802,7 +1802,6 @@ async def _ai_draft_reply_to_email(uid, folder="INBOX", reply_all=False, account
         from src.endpoint_resolver import (
             resolve_endpoint,
             resolve_utility_fallback_candidates,
-            resolve_chat_fallback_candidates,
         )
         from src.llm_core import llm_call_async_with_fallback
     except Exception as exc:
@@ -1843,13 +1842,6 @@ async def _ai_draft_reply_to_email(uid, folder="INBOX", reply_all=False, account
         utility_fallbacks = resolve_utility_fallback_candidates() or []
     for cand in utility_fallbacks:
         _add(*cand)
-    try:
-        chat_fallbacks = resolve_chat_fallback_candidates(owner=None) or []
-    except TypeError:
-        chat_fallbacks = resolve_chat_fallback_candidates() or []
-    for cand in chat_fallbacks:
-        _add(*cand)
-
     if not candidates:
         return {"error": "No LLM endpoint configured for AI reply"}
 

requirements-optional.txt

@@ -12,6 +12,16 @@
 # GPU-accelerated transcription — it's auto-detected, CPU is used otherwise.
 faster-whisper
 
+# Local text-to-speech via Kokoro-82M for the "local" TTS provider.
+# Kokoro 0.9.4 declares Python >=3.10,<3.13; Odysseus itself requires 3.11+,
+# so pip installs these extras on 3.11-3.12 and deliberately skips them on
+# Python 3.13+ (including the Python 3.14 container image). Kokoro declares
+# torch; the local provider still
+# requires a CUDA-enabled torch build and GPU at runtime. SoundFile is separate
+# in Kokoro's official install instructions and is not a transitive dependency.
+kokoro==0.9.4; python_version >= "3.11" and python_version < "3.13"
+soundfile; python_version >= "3.11" and python_version < "3.13"
+
 # DuckDuckGo as a search provider option.
 # Install if you want DDG in the search-provider dropdown.
 # Alternatives: SearXNG, Brave, Tavily, Serper, Google PSE.

routes/assistant_routes.py

@@ -16,7 +16,7 @@ from pydantic import BaseModel
 
 from core.database import SessionLocal, CrewMember, ScheduledTask
 from src.auth_helpers import get_current_user
-from core.auth import RESERVED_USERNAMES
+from src.owner_identity import REQUEST_SENTINEL_OWNERS
 from src.task_scheduler import compute_next_run
 
 
@@ -90,11 +90,12 @@ def setup_assistant_routes(task_scheduler) -> APIRouter:
     # check-in tasks seeded. Hitting any /assistant route under one of these
     # used to seed a full CrewMember + Morning/Midday/Evening tasks under that
     # owner, which then double-fired alongside the real user's check-ins.
-    # RESERVED_USERNAMES covers the same set; the `not owner` guard handles "".
+    # REQUEST_SENTINEL_OWNERS covers request-only identities; Default/Local is a
+    # reserved login name but remains a valid storage owner.
 
     async def _get_or_create(owner: str) -> CrewMember:
         """Return the per-owner assistant CrewMember, creating it on demand."""
-        if not owner or owner in RESERVED_USERNAMES:
+        if not owner or owner in REQUEST_SENTINEL_OWNERS:
             raise HTTPException(status_code=400, detail=f"Cannot seed assistant for {owner!r}")
         db = SessionLocal()
         try:

routes/auth_routes.py

@@ -22,6 +22,8 @@ from src.settings import (
     load_features as _load_features,
     save_features as _save_features,
     DEFAULT_SETTINGS,
+    RETIRED_SETTING_KEYS,
+    without_retired_settings,
 )
 from src.integrations import (
     load_integrations,
@@ -84,6 +86,33 @@ class SetOpenRegistrationRequest(BaseModel):
 SESSION_COOKIE = "odysseus_session"
 
 
+def _secure_cookie(request: Request) -> bool:
+    """Decide the ``Secure`` attribute of the session cookie.
+
+    ``SECURE_COOKIES`` stays authoritative when it holds an explicit value:
+    ``true`` always marks the cookie Secure (the documented knob for a TLS
+    proxy), ``false`` never does, which is the escape hatch for an install
+    that still answers on plain HTTP alongside HTTPS. Anything else —
+    unset, or the present-but-empty value docker-compose injects for a
+    variable the host has not defined — derives it from the request, so an
+    HTTPS login gets a Secure cookie without any configuration.
+
+    Either the connection scheme or ``X-Forwarded-Proto`` saying https is
+    enough, which is the same test ``core/middleware.py`` applies before it
+    sends HSTS. Uvicorn's proxy-headers middleware already folds that header
+    into the scheme for the proxies it trusts, so reading it here only adds
+    the case of a terminator that is not on a trusted address; the cost is
+    that a client talking to the app directly can set the header and lock
+    its own session out over plain HTTP.
+    """
+    configured = os.getenv("SECURE_COOKIES", "").strip().lower()
+    if configured in ("true", "false"):
+        return configured == "true"
+    # A chained proxy sends a list — the client-facing hop comes first.
+    forwarded_proto = request.headers.get("x-forwarded-proto", "").split(",")[0]
+    return request.url.scheme == "https" or forwarded_proto.strip().lower() == "https"
+
+
 def setup_auth_routes(auth_manager: AuthManager) -> APIRouter:
     router = APIRouter(prefix="/api/auth", tags=["auth"])
 
@@ -157,7 +186,7 @@ def setup_auth_routes(auth_manager: AuthManager) -> APIRouter:
             value=token,
             httponly=True,
             samesite="lax",
-            secure=os.getenv("SECURE_COOKIES", "false").lower() == "true",
+            secure=_secure_cookie(request),
             path="/",
         )
         if body.remember:
@@ -689,7 +718,7 @@ def setup_auth_routes(auth_manager: AuthManager) -> APIRouter:
         a scrubbed copy with secret keys blanked. The frontend uses this
         for keybinds + TTS prefs, so it stays callable without admin."""
         user = _get_current_user(request)
-        settings = _load_settings()
+        settings = without_retired_settings(_load_settings())
         if user and auth_manager.is_admin(user):
             return settings
         return scrub_settings(settings)
@@ -709,6 +738,8 @@ def setup_auth_routes(auth_manager: AuthManager) -> APIRouter:
             "agent_max_tool_calls": (0, 1000),  # 0 = unlimited
         }
         for key in DEFAULT_SETTINGS:
+            if key in RETIRED_SETTING_KEYS:
+                continue
             if key not in body:
                 continue
             val = body[key]
@@ -721,7 +752,7 @@ def setup_auth_routes(auth_manager: AuthManager) -> APIRouter:
                 val = max(lo, min(val, hi))
             current[key] = val
         _save_settings(current)
-        return current
+        return without_retired_settings(current)
 
     # ---- Integrations CRUD ----
 

routes/chat_helpers.py

@@ -15,7 +15,7 @@ from core.database import Session as DBSession, ModelEndpoint
 from src.llm_core import normalize_model_id
 from src.endpoint_resolver import normalize_base
 from src.context_compactor import maybe_compact, trim_for_context
-from src.model_context import estimate_tokens
+from src.model_context import estimate_tokens, get_context_length
 from src.auth_helpers import effective_user
 from src.prompt_security import untrusted_context_message
 from src.attachment_refs import attachment_ref
@@ -152,10 +152,38 @@ class ChatContext:
     # Uploads attached to this user turn, resolved and owner-checked for the
     # agent's private context. This is not emitted to the browser.
     uploaded_files: list = field(default_factory=list)
+    # Route-neutral prompt before any model-window compaction/trimming. This is
+    # retained only when explicit foreground fallbacks are enabled so each
+    # concrete candidate can apply its own context budget independently.
+    route_messages: list = field(default_factory=list)
 
 
 # ── Helpers ────────────────────────────────────────────────────────────── #
 
+def _allowed_models_from_privileges(privs: dict) -> Optional[frozenset[str]]:
+    if privs.get("block_all_models"):
+        return frozenset()
+    allowed_raw = privs.get("allowed_models")
+    allowed = allowed_raw if isinstance(allowed_raw, list) else []
+    restricted = bool(privs.get("allowed_models_restricted")) or bool(allowed)
+    return frozenset(model for model in allowed if isinstance(model, str)) if restricted else None
+
+
+def _allowed_models_for_request(request) -> Optional[frozenset[str]]:
+    """Return the caller's model allowlist, or ``None`` when unrestricted."""
+
+    try:
+        user = effective_user(request)
+    except Exception:
+        user = None
+    if not user:
+        return None
+    auth_manager = getattr(getattr(request.app, "state", None), "auth_manager", None)
+    if not auth_manager:
+        return None
+    privs = auth_manager.get_privileges(user) or {}
+    return _allowed_models_from_privileges(privs)
+
 def _enforce_chat_privileges(request, sess) -> None:
     """Apply the per-user privilege gates (allowed_models + max_messages_per_day)
     that both /api/chat and /api/chat_stream must enforce BEFORE any LLM work.
@@ -185,10 +213,8 @@ def _enforce_chat_privileges(request, sess) -> None:
     if privs.get("block_all_models"):
         raise HTTPException(403, f"Your account is not allowed to use model '{sess.model}'.")
 
-    allowed_raw = privs.get("allowed_models")
-    allowed = allowed_raw if isinstance(allowed_raw, list) else []
-    restricted = bool(privs.get("allowed_models_restricted")) or bool(allowed)
-    if restricted and sess.model and sess.model not in allowed:
+    allowed_models = _allowed_models_from_privileges(privs)
+    if allowed_models is not None and sess.model and sess.model not in allowed_models:
         raise HTTPException(403, f"Your account is not allowed to use model '{sess.model}'.")
 
     cap = int(privs.get("max_messages_per_day") or 0)
@@ -287,96 +313,6 @@ async def auto_name_session(session_manager, sess):
         logger.error(f"Auto-name failed for {sess.id}: {e}\n{traceback.format_exc()}")
 
 
-def try_fallback_endpoint(sess, session_id: str) -> dict | None:
-    """Find an alternative working endpoint when the current one fails.
-
-    Returns {"model": ..., "endpoint_url": ..., "endpoint_name": ...} or None.
-    """
-    import requests as _req
-    from src.endpoint_resolver import (
-        build_chat_url,
-        build_headers,
-        build_models_url,
-        normalize_base,
-        resolve_endpoint_runtime,
-    )
-    from src.chatgpt_subscription import is_chatgpt_subscription_base
-
-    current_url = sess.endpoint_url or ""
-    owner = getattr(sess, "owner", None)
-    db = SessionLocal()
-    try:
-        q = db.query(ModelEndpoint).filter(
-            ModelEndpoint.is_enabled == True
-        )
-        if owner:
-            from src.auth_helpers import owner_filter
-            q = owner_filter(q, ModelEndpoint, owner)
-        endpoints = q.all()
-    finally:
-        db.close()
-
-    for ep in endpoints:
-        base = normalize_base(ep.base_url)
-        # Skip current endpoint
-        if current_url and base in current_url:
-            continue
-        try:
-            base, api_key = resolve_endpoint_runtime(ep, owner=owner)
-        except Exception:
-            continue
-        ping_url = build_models_url(base)
-        headers = build_headers(api_key, base)
-        try:
-            if ping_url:
-                r = _req.get(ping_url, headers=headers, timeout=5)
-                r.raise_for_status()
-                data = r.json()
-                models = [m.get("id") for m in (data.get("data") or []) if m.get("id")]
-                if not models:
-                    models = [
-                        m.get("name") or m.get("model")
-                        for m in (data.get("models") or [])
-                        if m.get("name") or m.get("model")
-                    ]
-            else:
-                models = json.loads(ep.cached_models or "[]")
-            if not models:
-                continue
-            # Found a working endpoint — update session
-            new_model = models[0]
-            chat_url = build_chat_url(base)
-            new_headers = build_headers(api_key, base)
-            persisted_headers = {} if is_chatgpt_subscription_base(base) else new_headers
-
-            sess.model = new_model
-            sess.endpoint_url = chat_url
-            sess.headers = new_headers
-
-            # Persist
-            _db = SessionLocal()
-            try:
-                _db.query(DBSession).filter(DBSession.id == session_id).update({
-                    "model": new_model,
-                    "endpoint_url": chat_url,
-                    "headers": persisted_headers,
-                })
-                _db.commit()
-            finally:
-                _db.close()
-
-            logger.info(f"Fallback: switched session {session_id} from {current_url} to {ep.name} ({new_model})")
-            return {
-                "model": new_model,
-                "endpoint_url": chat_url,
-                "endpoint_name": ep.name,
-            }
-        except Exception:
-            continue
-
-    return None
-
-
 def extract_preset(chat_handler, preset_id) -> PresetInfo:
     """Extract preset parameters via chat_handler."""
     temperature, max_tokens, system_prompt, char_name = (
@@ -687,6 +623,7 @@ async def build_chat_context(
     use_enhanced_message: bool = False,
     agent_mode: bool = False,
     allow_tool_preprocessing: bool = True,
+    defer_context_shaping: bool = False,
 ) -> ChatContext:
     """Build the full context (preface + messages) for an LLM call.
 
@@ -830,13 +767,22 @@ async def build_chat_context(
         except Exception:
             logger.debug("Failed to add current date/time context", exc_info=True)
 
-    # Auto-compact
-    messages, context_length, was_compacted = await maybe_compact(
-        sess, sess.endpoint_url, sess.model, messages, sess.headers, owner=user,
-    )
+    route_messages = list(messages)
+    # Explicit fallback routing must shape from the same route-neutral prompt
+    # for every candidate. Running selected-model compaction here would mutate
+    # session history before we know which route can answer and would make a
+    # later larger-context candidate unable to recover discarded history.
+    if defer_context_shaping:
+        context_length = get_context_length(sess.endpoint_url, sess.model)
+        was_compacted = False
+    else:
+        messages, context_length, was_compacted = await maybe_compact(
+            sess, sess.endpoint_url, sess.model, messages, sess.headers, owner=user,
+        )
     _before_trim_messages = len(messages)
     _before_trim_tokens = estimate_tokens(messages)
-    messages = trim_for_context(messages, context_length)
+    if not defer_context_shaping:
+        messages = trim_for_context(messages, context_length)
     _after_trim_messages = len(messages)
     _after_trim_tokens = estimate_tokens(messages)
     _context_trimmed = _after_trim_messages < _before_trim_messages or _after_trim_tokens < _before_trim_tokens
@@ -860,6 +806,7 @@ async def build_chat_context(
         context_tokens_after_trim=_after_trim_tokens,
         auto_opened_docs=auto_opened_docs,
         uploaded_files=uploaded_files,
+        route_messages=route_messages,
     )
 
 

routes/email_routes.py

@@ -5004,7 +5004,6 @@ def setup_email_routes():
             from src.endpoint_resolver import (
                 resolve_endpoint,
                 resolve_utility_fallback_candidates,
-                resolve_chat_fallback_candidates,
             )
             from src.llm_core import llm_call_async_with_fallback
 
@@ -5066,8 +5065,6 @@ def setup_email_routes():
                 pass
             for cand in resolve_utility_fallback_candidates(owner=owner) or []:
                 _add(*cand)
-            for cand in resolve_chat_fallback_candidates(owner=owner) or []:
-                _add(*cand)
             if not candidates:
                 return {"success": False, "error": "No LLM endpoint configured"}
 
@@ -5327,13 +5324,11 @@ def setup_email_routes():
             # Build a candidate chain so a stale session-stored API key
             # (the most common cause of "authentication failed" here)
             # doesn't kill AI Reply outright — fall through to the
-            # user's Utility / Default endpoints and the active Utility
-            # fallback chain. The retired default-fallback hook stays empty.
-            # Dedupe by url+model so we don't retry the same broken endpoint.
+            # user's Utility / Default endpoints and active Utility fallback
+            # chain. Dedupe by url+model so we don't retry the same endpoint.
             from src.llm_core import llm_call_async_with_fallback
             from src.endpoint_resolver import (
                 resolve_utility_fallback_candidates,
-                resolve_chat_fallback_candidates,
             )
             _seen = set()
             _candidates = []
@@ -5358,11 +5353,9 @@ def setup_email_routes():
                 _add(_d_url, _d_model, _d_headers)
             except Exception:
                 pass
-            # Active Utility fallbacks, then the retired default hook.
+            # Active Utility fallbacks last.
             for cand in resolve_utility_fallback_candidates(owner=owner) or []:
                 _add(*cand)
-            for cand in resolve_chat_fallback_candidates(owner=owner) or []:
-                _add(*cand)
             _messages = [
                 {"role": "system", "content": system_prompt},
                 {"role": "user", "content": user_msg},

routes/mcp/mcp_routes.py

@@ -475,7 +475,7 @@ def setup_mcp_routes(mcp_manager: McpManager):
                 return RedirectResponse(auth_url)
             else:
                 # Remote device — show paste-back page
-                return HTMLResponse(_oauth_authorize_page(auth_url, server_id, host, redirect_uri))
+                return HTMLResponse(_oauth_authorize_page(auth_url, server_id, redirect_uri))
         finally:
             db.close()
 
@@ -612,15 +612,13 @@ def setup_mcp_routes(mcp_manager: McpManager):
 def _oauth_authorize_page(
     auth_url: str,
     server_id: str,
-    host: str,
-    redirect_uri: str = "http://localhost:7000/api/mcp/oauth/callback",
+    redirect_uri: str,
 ) -> str:
     """Page with Google sign-in link and URL paste-back form for remote access."""
-    # Escape values interpolated into the page: `host` comes from the request
-    # Host header and `server_id` from the OAuth state — neither is trusted.
+    # Escape values interpolated into the page: `server_id` comes from the OAuth
+    # state and is not trusted.
     auth_url = html.escape(auth_url, quote=True)
     server_id = html.escape(server_id, quote=True)
-    host = html.escape(host, quote=True)
     redirect_uri = html.escape(redirect_uri, quote=True)
     return f"""<!DOCTYPE html>
 <html><head>
@@ -664,7 +662,15 @@ def _oauth_authorize_page(
   </div>
   <a class="auth-link" href="{auth_url}" target="_blank" rel="noopener">Sign in with Google</a>
   <div class="divider"></div>
-  <form method="POST" action="http://{host}/api/mcp/oauth/exchange/{server_id}">
+  <!-- Relative action: the browser resolves it against the origin this page was
+       served from, so the form follows the user through any proxy without the
+       app having to know the scheme or the host. An absolute http:// action is
+       blocked as mixed content on exactly the HTTPS deployments that need
+       paste-back, and request.url.scheme cannot be trusted to spot them —
+       uvicorn only honours X-Forwarded-Proto from a peer in
+       --forwarded-allow-ips, which defaults to 127.0.0.1 and excludes a proxy
+       arriving over the Docker bridge. -->
+  <form method="POST" action="/api/mcp/oauth/exchange/{server_id}">
     <p>Paste the URL from your browser after signing in:</p>
     <input type="text" name="callback_url" placeholder="{redirect_uri}?code=..." required>
     <br><button type="submit">Connect</button>

routes/model_routes.py

@@ -46,6 +46,7 @@ _ENDPOINT_SETTING_FIELDS = {
 }
 
 _ENDPOINT_FALLBACK_FIELDS = {
+    "foreground_model_fallbacks": "Foreground Model Fallbacks",
     "utility_model_fallbacks": "Utility Model Fallbacks",
     "vision_model_fallbacks":  "Vision Model Fallbacks",
 }
@@ -180,7 +181,12 @@ def _clear_user_pref_endpoint_refs(all_prefs: dict, ep_id: str) -> int:
     if not isinstance(all_prefs, dict):
         return 0
     users = all_prefs.get("_users")
-    pref_sets = users.values() if isinstance(users, dict) else [all_prefs]
+    # A mixed store can contain auth-disabled foreground policy at the root
+    # alongside named-owner preferences. Both are active namespaces; legacy
+    # `default_model_fallbacks` remains untouched by the field allowlist.
+    pref_sets = [all_prefs]
+    if isinstance(users, dict):
+        pref_sets.extend(users.values())
     cleared_users = 0
     for prefs in pref_sets:
         if isinstance(prefs, dict) and _clear_endpoint_settings_for_endpoint(prefs, ep_id):
@@ -1345,14 +1351,14 @@ def _legacy_visible_api_models(ep) -> List[str]:
 def _picker_models_for_endpoint(ep, base_url: str, kind: str):
     """Return model IDs that should appear in the picker for an endpoint.
 
-    API providers expose remote inventory from /v1/models. Treat that cache as
-    inventory, not approval: only manually pinned API models should appear in
-    the picker. Local/self-hosted endpoints keep the older hide-list behavior.
+    API providers expose remote inventory from /v1/models. Default to that
+    visible inventory until an explicit pinned-model allow-list is saved.
+    Local/self-hosted endpoints keep the older hide-list behavior.
     """
     pinned = _normalize_model_ids(getattr(ep, "pinned_models", None))
     if _picker_requires_pinning(base_url, kind):
         if not _has_explicit_pinned_models(ep):
-            pinned = _legacy_visible_api_models(ep) if _hidden_model_ids(ep) else []
+            pinned = _legacy_visible_api_models(ep)
         return pinned, pinned
     return _visible_models(
         _cached_model_ids(ep),
@@ -2336,9 +2342,7 @@ def setup_model_routes(model_discovery):
                 else:
                     response.headers["X-Model-Refresh-Status"] = "failed"
                     response.headers["X-Model-Refresh-Warning"] = "Model refresh failed or returned no models; kept cached models."
-            pinned = _normalize_model_ids(getattr(ep, "pinned_models", None))
-            if picker_requires_pinning and not _has_explicit_pinned_models(ep):
-                pinned = _legacy_visible_api_models(ep)
+            _, pinned = _picker_models_for_endpoint(ep, base, kind)
             pinned_set = set(pinned)
             return [
                 {

routes/prefs_routes.py

@@ -7,6 +7,10 @@ from src.auth_helpers import get_current_user
 from src.constants import USER_PREFS_FILE
 
 PREFS_FILE = USER_PREFS_FILE
+_FOREGROUND_POLICY_KEYS = (
+    "foreground_fallback_enabled",
+    "foreground_model_fallbacks",
+)
 
 
 def _load():
@@ -26,14 +30,27 @@ def _save(prefs):
 def _load_for_user(user: Optional[str] = None) -> dict:
     """Load preferences for a specific user."""
     all_prefs = _load()
-    if "_users" in all_prefs:
+    users = all_prefs.get("_users")
+    if isinstance(users, dict):
         if user is None:
             # Auth disabled — return first user's prefs for backward compat
-            users = all_prefs["_users"]
-            return dict(next(iter(users.values()), {}))
-        return dict(all_prefs["_users"].get(user, {}))
-    # Legacy flat format — return as-is
-    return dict(all_prefs)
+            prefs = dict(next(iter(users.values()), {}))
+            # Foreground fallback consent is never borrowed from a named
+            # owner. Auth-disabled operation has a separate flat/root opt-in
+            # that remains inert when authentication is enabled again.
+            for key in _FOREGROUND_POLICY_KEYS:
+                prefs.pop(key, None)
+                if key in all_prefs:
+                    prefs[key] = all_prefs[key]
+            return prefs
+        prefs = users.get(user, {})
+        return dict(prefs) if isinstance(prefs, dict) else {}
+    # A legacy flat store belongs only to auth-disabled single-user mode.
+    # Copying it into the first named user's new `_users` record during an
+    # auth transition would silently transfer another user's preferences and,
+    # critically, foreground fallback consent. Named owners therefore start
+    # with an empty record and must write their own preferences explicitly.
+    return dict(all_prefs) if user is None else {}
 
 
 def _save_for_user(user: Optional[str], prefs: dict):
@@ -45,17 +62,40 @@ def _save_for_user(user: Optional[str], prefs: dict):
         # `prefs` flat would overwrite the whole `_users` map and destroy every
         # other user's preferences. Instead write back into the same (first)
         # slot _load_for_user(None) reads from, preserving the others.
-        if "_users" in all_prefs:
-            users = all_prefs["_users"]
+        users = all_prefs.get("_users")
+        if isinstance(users, dict):
             first_key = next(iter(users), None)
             if first_key is not None:
-                users[first_key] = prefs
+                existing_named = users.get(first_key)
+                existing_named = (
+                    dict(existing_named)
+                    if isinstance(existing_named, dict)
+                    else {}
+                )
+                named_foreground = {
+                    key: existing_named[key]
+                    for key in _FOREGROUND_POLICY_KEYS
+                    if key in existing_named
+                }
+                users[first_key] = {
+                    key: value
+                    for key, value in prefs.items()
+                    if key not in _FOREGROUND_POLICY_KEYS
+                }
+                users[first_key].update(named_foreground)
+                for key in _FOREGROUND_POLICY_KEYS:
+                    if key in prefs:
+                        all_prefs[key] = prefs[key]
                 _save(all_prefs)
                 return
         _save(prefs)
         return
-    if "_users" not in all_prefs:
-        all_prefs = {"_users": {}}
+    if not isinstance(all_prefs.get("_users"), dict):
+        # Preserve the flat single-user object as inert legacy data while
+        # creating the first named-owner namespace. In particular, historical
+        # fallback values must not be deleted or copied into the new owner.
+        all_prefs = dict(all_prefs)
+        all_prefs["_users"] = {}
     all_prefs["_users"][user] = prefs
     _save(all_prefs)
 

routes/research/research_routes.py

@@ -15,7 +15,7 @@ from pydantic import BaseModel, Field
 from core.middleware import INTERNAL_TOOL_USER
 from src.endpoint_resolver import resolve_endpoint
 from src.auth_helpers import _auth_disabled, get_current_user
-from core.auth import RESERVED_USERNAMES
+from src.owner_identity import REQUEST_SENTINEL_OWNERS
 from src.constants import DEEP_RESEARCH_DIR
 
 _SESSION_ID_RE = re.compile(r"^[a-zA-Z0-9-]{1,128}$")
@@ -496,7 +496,7 @@ def setup_research_routes(research_handler, session_manager=None) -> APIRouter:
         user = require_privilege(request, "can_use_research")
         if user == INTERNAL_TOOL_USER:
             tool_owner = (request.headers.get("X-Odysseus-Owner") or "").strip()
-            if tool_owner and tool_owner not in RESERVED_USERNAMES:
+            if tool_owner and tool_owner not in REQUEST_SENTINEL_OWNERS:
                 auth_mgr = getattr(request.app.state, "auth_manager", None)
                 if auth_mgr is not None and getattr(auth_mgr, "is_configured", False):
                     try:

routes/skills_routes.py

@@ -18,6 +18,7 @@ from pydantic import BaseModel, Field
 
 from services.memory.skills import SkillsManager
 from src.auth_helpers import get_current_user
+from src.prompt_security import untrusted_context_message
 from core.middleware import require_admin
 
 logger = logging.getLogger(__name__)
@@ -107,6 +108,23 @@ def _skill_test_task(skill: dict) -> str:
     )
 
 
+def _skill_test_messages(md: str, task: str) -> list[dict]:
+    """Keep user-editable skill text out of the trusted system role."""
+    return [
+        {
+            "role": "system",
+            "content": (
+                "You are TESTING a skill. Follow the supplied reusable procedure "
+                "to complete the user's task for real, using available tools step "
+                "by step. If the skill is wrong, unclear, or references tools that "
+                "do not exist, do your best; the problems will be reviewed afterward."
+            ),
+        },
+        untrusted_context_message("skill under test", md),
+        {"role": "user", "content": task},
+    ]
+
+
 async def _eval_skill_run(skill_md: str, task: str, transcript: str,
                           url: str, model: str, headers: Optional[dict]) -> dict:
     """LLM-as-judge: grade a skill test run from its transcript. Advisory only.
@@ -411,7 +429,21 @@ async def _eval_skill_retrieval_precision(skill_md: str, others: list,
 _skill_test_jobs: dict = {}
 
 
-async def _run_skill_test_job(key, name, md, task, url, model, headers, owner, skills_manager=None):
+async def _run_skill_test_job(
+    key,
+    name,
+    md,
+    task,
+    url,
+    model,
+    headers,
+    owner,
+    skills_manager=None,
+    *,
+    messages=None,
+    transcript=None,
+    exact_approval=None,
+):
     """Background coroutine: run the skill in an agent loop, capture a condensed
     log + transcript, then have the judge grade it. Writes into _skill_test_jobs."""
     import json as _json
@@ -421,7 +453,7 @@ async def _run_skill_test_job(key, name, md, task, url, model, headers, owner, s
     if job is None:
         return
     log = job["log"]
-    transcript = []
+    transcript = transcript if isinstance(transcript, list) else []
     say_buf = []
 
     def _flush_say():
@@ -429,18 +461,12 @@ async def _run_skill_test_job(key, name, md, task, url, model, headers, owner, s
             log.append({"type": "say", "text": "".join(say_buf)})
             say_buf.clear()
 
-    messages = [
-        {"role": "system", "content":
-            "You are TESTING a skill. Below is a reusable skill (a procedure). Follow it "
-            "to complete the user's task for real, using your available tools, step by "
-            "step. If the skill is wrong, unclear, or references tools that don't exist, "
-            "do your best — the problems will be reviewed afterward.\n\n=== SKILL ===\n" + md},
-        {"role": "user", "content": task},
-    ]
+    messages = list(messages) if isinstance(messages, list) else _skill_test_messages(md, task)
     try:
         async for chunk in stream_agent_loop(
             url, model, messages, headers=headers,
             temperature=0.3, max_tokens=0, max_rounds=8, owner=owner,
+            exact_approval=exact_approval,
         ):
             if not chunk.startswith("data: ") or chunk.strip() == "data: [DONE]":
                 continue
@@ -458,8 +484,25 @@ async def _run_skill_test_job(key, name, md, task, url, model, headers, owner, s
             elif d.get("type") == "tool_output":
                 _flush_say()
                 out = str(d.get("output") or "")[:600]
-                log.append({"type": "tool_output", "output": out})
+                tool_log = {"type": "tool_output", "output": out}
+                approval = d.get("ask_user")
+                if isinstance(approval, dict):
+                    tool_log["ask_user"] = approval
+                log.append(tool_log)
                 transcript.append(f"[output] {out}\n")
+                if (
+                    isinstance(approval, dict)
+                    and approval.get("kind") == "tool_approval"
+                    and approval.get("approval_id")
+                ):
+                    # Manual skill tests have their own polling UI instead of a
+                    # chat session. Pause the run and retain only server-side
+                    # continuation state until the same owner approves/denies
+                    # this exact sealed action.
+                    job["status"] = "awaiting_approval"
+                    job["approval"] = approval
+                    job["_transcript"] = transcript
+                    return
             elif d.get("type") == "agent_step":
                 _flush_say()
                 log.append({"type": "agent_step", "round": d.get("round")})
@@ -471,6 +514,9 @@ async def _run_skill_test_job(key, name, md, task, url, model, headers, owner, s
         _flush_say()
         log.append({"type": "error", "error": str(e)})
 
+    job.pop("approval", None)
+    job.pop("_transcript", None)
+    job.pop("_run", None)
     log.append({"type": "evaluating"})
     try:
         job["verdict"] = await _eval_skill_run(md, task, "".join(transcript), url, model, headers)
@@ -694,12 +740,8 @@ async def _run_skill_test_once(md: str, task: str, url, model, headers, owner) -
     import json as _json
     from src.agent_loop import stream_agent_loop
     transcript = []
-    messages = [
-        {"role": "system", "content":
-            "You are TESTING a skill. Follow this skill's procedure to complete the task "
-            "for real, using your tools, step by step.\n\n=== SKILL ===\n" + md},
-        {"role": "user", "content": task},
-    ]
+    approval_required = None
+    messages = _skill_test_messages(md, task)
     try:
         # max_tokens explicitly set: passing 0 lets some upstreams (Ollama,
         # OpenAI-compat) generate an empty completion, which manifested as
@@ -719,11 +761,44 @@ async def _run_skill_test_once(md: str, task: str, url, model, headers, owner) -
                 transcript.append(f"\n[tool {d.get('tool')}] {str(d.get('command') or d.get('args') or '')[:300]}\n")
             elif d.get("type") == "tool_output":
                 transcript.append(f"[output] {str(d.get('output') or '')[:600]}\n")
+                approval = d.get("ask_user")
+                if (
+                    isinstance(approval, dict)
+                    and approval.get("kind") == "tool_approval"
+                ):
+                    approval_required = approval
+                    break
             elif d.get("type") == "agent_step":
                 transcript.append(f"\n--- round {d.get('round')} ---\n")
     except Exception as e:
         transcript.append(f"\n[run error] {e}\n")
     text = "".join(transcript)
+    if approval_required is not None:
+        # Unattended audits have no authority to approve and no UI that could
+        # resume this record. Destructively deny it now instead of leaving a
+        # reusable opaque grant pending until TTL/cap eviction.
+        try:
+            from src.tool_approvals import tool_approval_store
+            tool_approval_store.consume(
+                approval_required.get("approval_id"),
+                decision="deny",
+                owner=owner,
+                session_id=None,
+            )
+        except Exception:
+            logger.debug("Could not retire unattended skill approval", exc_info=True)
+        return text, {
+            "verdict": "inconclusive",
+            "confidence": 1.0,
+            "summary": (
+                "This automated audit reached an exact action that requires "
+                "a human approval; no action was executed."
+            ),
+            "issues": [
+                "Run this skill's manual test and review the sealed action."
+            ],
+            "approval_required": True,
+        }
     verdict = await _eval_skill_run(md, task, text, url, model, headers)
     return text, verdict
 
@@ -863,6 +938,26 @@ async def _audit_one_skill(skills_manager, skill, url, model, headers,
     transcript, verdict = await _run_skill_test_once(md, task, url, model, headers, owner)
     v = verdict.get("verdict")
     log(f"{name}: verdict = {v} ({verdict.get('summary', '')[:80]})")
+    if verdict.get("approval_required"):
+        # An unattended audit is not authority for an action influenced by the
+        # skill under test. Preserve the skill's current publication/confidence
+        # state and route the exact action to the manual test UI instead of
+        # letting a safety pause demote, rewrite, or auto-publish the skill.
+        skills_manager.set_audit(
+            name,
+            "inconclusive",
+            by_teacher=False,
+            worker_model=model,
+            owner=owner,
+        )
+        status = skill.get("status") or "draft"
+        log(f"{name}: {status} unchanged — exact action needs manual approval")
+        return {
+            "skill": name,
+            "result": "approval_required",
+            "verdict": verdict,
+            "status": status,
+        }
     if v == "pass":
         # Procedure works. If the reviewer still flagged metadata (tags/category/
         # when_to_use/description), do ONE fixer pass to correct the frontmatter
@@ -1431,6 +1526,19 @@ def setup_skills_routes(skills_manager: SkillsManager) -> APIRouter:
             logger.warning(f"Skill-test model resolve failed: {_e}")
 
         key = (user or "", name)
+        previous_job = _skill_test_jobs.get(key) or {}
+        previous_approval = previous_job.get("approval") or {}
+        if previous_approval.get("approval_id"):
+            try:
+                from src.tool_approvals import tool_approval_store
+                tool_approval_store.consume(
+                    previous_approval["approval_id"],
+                    decision="deny",
+                    owner=user,
+                    session_id=None,
+                )
+            except Exception:
+                logger.debug("Could not retire replaced skill approval", exc_info=True)
         _skill_test_jobs[key] = {
             "status": "running",
             "task": task,
@@ -1439,10 +1547,135 @@ def setup_skills_routes(skills_manager: SkillsManager) -> APIRouter:
             "started": _time.time(),
             "log": [{"type": "skill_test_start", "task": task, "skill": name, "model": model}],
             "verdict": None,
+            "_run": {
+                "md": md,
+                "url": url,
+                "model": model,
+                "headers": headers,
+                "owner": user,
+            },
         }
         _asyncio.create_task(_run_skill_test_job(key, name, md, task, url, model, headers, user, skills_manager))
         return {"ok": True, "status": "running", "skill": name, "model": model}
 
+    @router.post("/{skill_id}/test-approval")
+    async def approve_skill_test_action(request: Request, skill_id: str):
+        """Resume a manual skill test with one exact server-sealed action."""
+        import asyncio as _asyncio
+        from src.tool_approvals import tool_approval_store
+
+        user = _owner(request)
+        skills = skills_manager.load(owner=user)
+        match = next(
+            (s for s in skills if s.get("name") == skill_id or s.get("id") == skill_id),
+            None,
+        )
+        if not match:
+            raise HTTPException(404, "Skill not found")
+        _verify_owner(match, user)
+        name = match.get("name")
+        key = (user or "", name)
+        job = _skill_test_jobs.get(key)
+        if not job or job.get("status") != "awaiting_approval":
+            raise HTTPException(409, "This skill test is not awaiting an approval.")
+
+        body = await request.json()
+        if not isinstance(body, dict):
+            raise HTTPException(400, "Tool approval body must be a JSON object.")
+        approval_id = str(body.get("approval_id") or "")
+        decision = str(body.get("decision") or "").strip().lower()
+        expected = job.get("approval") or {}
+        if approval_id != str(expected.get("approval_id") or ""):
+            raise HTTPException(409, "This approval does not match the pending skill test action.")
+        if decision not in {"approve", "deny"}:
+            raise HTTPException(400, "Invalid tool approval decision.")
+
+        pending = tool_approval_store.peek(approval_id)
+        normalized_owner = str(user or "").strip().casefold()
+        if (
+            pending is None
+            or pending.owner != normalized_owner
+            or pending.session_id != ""
+        ):
+            raise HTTPException(409, "This tool approval is invalid or expired.")
+        exact_approval = tool_approval_store.consume(
+            approval_id,
+            decision=decision,
+            owner=user,
+            session_id=None,
+        )
+
+        if decision == "approve" and exact_approval is None:
+            raise HTTPException(409, "This tool approval could not be consumed.")
+        job.pop("approval", None)
+        if decision == "deny":
+            job.pop("_transcript", None)
+            job.pop("_run", None)
+            job["log"].append({
+                "type": "approval_denied",
+                "text": "Exact action denied; the skill test stopped without executing it.",
+            })
+            job["verdict"] = {
+                "verdict": "inconclusive",
+                "confidence": 1.0,
+                "summary": "The test stopped because its exact action was denied.",
+                "issues": [],
+            }
+            job["status"] = "done"
+            return {"ok": True, "status": "done", "decision": "deny"}
+
+        run = job.get("_run") or {}
+        transcript = job.pop("_transcript", [])
+        # stream_agent_loop owns its per-round message list internally. Rebuild
+        # continuation context from the original untrusted skill plus the
+        # accumulated transcript so repeated approvals do not lose earlier
+        # approved results, while keeping every transcript byte tainted.
+        messages = _skill_test_messages(
+            run.get("md", ""),
+            job.get("task", ""),
+        )
+        if transcript:
+            messages.append(untrusted_context_message(
+                "skill test transcript",
+                "".join(str(item) for item in transcript),
+            ))
+        messages.extend([
+            {
+                "role": "assistant",
+                "content": str(expected.get("question") or "Allow this exact action once?"),
+            },
+            {
+                "role": "user",
+                "content": (
+                    f"Approved the exact {exact_approval.pending.tool_name} "
+                    "action shown above once."
+                ),
+            },
+        ])
+        job["status"] = "running"
+        job["log"].append({
+            "type": "approval_granted",
+            "text": (
+                f"Approved exact {exact_approval.pending.tool_name} action once; "
+                "resuming test."
+            ),
+        })
+        _asyncio.create_task(_run_skill_test_job(
+            key,
+            name,
+            run.get("md", ""),
+            job.get("task", ""),
+            run.get("url"),
+            run.get("model"),
+            run.get("headers"),
+            run.get("owner"),
+            skills_manager,
+            messages=messages,
+            transcript=transcript,
+            exact_approval=exact_approval,
+        ))
+        return {"ok": True, "status": "running", "decision": "approve"}
+
     @router.get("/{skill_id}/test-status")
     async def test_skill_status(request: Request, skill_id: str):
         """Current background-test state for a skill (status / log / verdict)."""
@@ -1459,6 +1692,7 @@ def setup_skills_routes(skills_manager: SkillsManager) -> APIRouter:
             "model": job.get("model"),
             "log": job.get("log", []),
             "verdict": job.get("verdict"),
+            "approval": job.get("approval"),
         }
 
     @router.post("/audit-all")

routes/task/__init__.py

@@ -0,0 +1,5 @@
+"""Task route domain package (slice 2p, #4082/#4071).
+
+Contains task_routes.py, migrated from the flat routes/ directory.
+Backward-compat shim at routes/task_routes.py re-exports from here.
+"""

scripts/migrate_searxng_settings.py

@@ -0,0 +1,166 @@
+#!/usr/bin/env python3
+"""Make retained SearXNG settings inherit defaults without replacing them."""
+
+from __future__ import annotations
+
+import os
+import stat
+import sys
+import tempfile
+from pathlib import Path
+
+import yaml
+from yaml.nodes import MappingNode
+from yaml.tokens import BlockMappingStartToken, FlowMappingStartToken
+
+
+_UTF8_BOM = b"\xef\xbb\xbf"
+
+
+def _parse_root_mapping(text: str) -> tuple[MappingNode | None, dict]:
+    """Parse settings with the same safe YAML semantics SearXNG uses."""
+    try:
+        loaded = yaml.safe_load(text)
+        node = yaml.compose(text, Loader=yaml.SafeLoader)
+    except yaml.YAMLError:
+        raise ValueError("settings file is not valid single-document YAML") from None
+
+    if loaded is None and node is None:
+        return None, {}
+    if not isinstance(loaded, dict) or not isinstance(node, MappingNode):
+        raise ValueError("settings root is not a mapping")
+    return node, loaded
+
+
+def _flow_mapping_start(text: str) -> int:
+    """Return the root flow mapping's opening-brace character offset."""
+    try:
+        for token in yaml.scan(text, Loader=yaml.SafeLoader):
+            if isinstance(token, FlowMappingStartToken):
+                return token.start_mark.index
+    except yaml.YAMLError:
+        pass
+    raise ValueError("flow-style settings mapping has no opening brace")
+
+
+def _newline_for(contents: bytes) -> bytes:
+    first_lf = contents.find(b"\n")
+    if first_lf > 0 and contents[first_lf - 1 : first_lf + 1] == b"\r\n":
+        return b"\r\n"
+    return b"\n"
+
+
+def _block_mapping_position(text: str, root: MappingNode | None) -> tuple[int, int]:
+    """Return a safe character offset and indent for a root block mapping key."""
+    if root is None:
+        return len(text), 0
+
+    try:
+        for token in yaml.scan(text, Loader=yaml.SafeLoader):
+            if not isinstance(token, BlockMappingStartToken):
+                continue
+            line_start = token.start_mark.index - token.start_mark.column
+            if not text[line_start : token.start_mark.index].strip():
+                return line_start, token.start_mark.column
+            return root.end_mark.index, token.start_mark.column
+    except yaml.YAMLError:
+        pass
+    return root.end_mark.index, root.start_mark.column
+
+
+def _add_block_default_inheritance(
+    contents: bytes, text: str, root: MappingNode | None
+) -> bytes:
+    newline = _newline_for(contents)
+    character_offset, indent_width = _block_mapping_position(text, root)
+    bom_length = len(_UTF8_BOM) if contents.startswith(_UTF8_BOM) else 0
+    offset = bom_length + len(text[:character_offset].encode("utf-8"))
+    separator = b""
+    if offset not in (0, bom_length) and not contents[:offset].endswith((b"\n", b"\r")):
+        separator = newline
+    addition = (
+        separator
+        + b" " * indent_width
+        + b"use_default_settings: true"
+        + newline
+    )
+    return contents[:offset] + addition + contents[offset:]
+
+
+def migrate_settings(path: Path) -> bool:
+    """Add the missing inheritance key atomically; return whether the file changed."""
+    source_stat = path.lstat()
+    if not stat.S_ISREG(source_stat.st_mode):
+        raise ValueError(f"settings path is not a regular file: {path}")
+
+    contents = path.read_bytes()
+    if not contents:
+        return False
+
+    text = contents.decode("utf-8-sig")
+    root, loaded = _parse_root_mapping(text)
+    if "use_default_settings" in loaded:
+        return False
+
+    if root is not None and root.flow_style:
+        start = _flow_mapping_start(text)
+        bom_length = len(_UTF8_BOM) if contents.startswith(_UTF8_BOM) else 0
+        offset = bom_length + len(text[: start + 1].encode("utf-8"))
+        separator = b", " if root.value else b""
+        updated = (
+            contents[:offset]
+            + b"use_default_settings: true"
+            + separator
+            + contents[offset:]
+        )
+    else:
+        updated = _add_block_default_inheritance(contents, text, root)
+    fd, temporary_name = tempfile.mkstemp(
+        prefix=f".{path.name}.odysseus-", dir=path.parent
+    )
+    temporary = Path(temporary_name)
+    try:
+        # chmod before chown: the Compose cap set is `cap_drop: ALL` plus
+        # CHOWN/SETGID/SETUID/DAC_OVERRIDE, with no FOWNER. Once the temporary
+        # file belongs to searxng:searxng — which every retained settings file
+        # does, because searxng's entrypoint chowns /etc/searxng — root can no
+        # longer chmod it and the migration dies with EPERM.
+        os.fchmod(fd, stat.S_IMODE(source_stat.st_mode))
+        os.fchown(fd, source_stat.st_uid, source_stat.st_gid)
+        with os.fdopen(fd, "wb") as handle:
+            fd = -1
+            handle.write(updated)
+            handle.flush()
+            os.fsync(handle.fileno())
+        os.replace(temporary, path)
+        directory_fd = os.open(path.parent, os.O_RDONLY | os.O_DIRECTORY)
+        try:
+            os.fsync(directory_fd)
+        finally:
+            os.close(directory_fd)
+    finally:
+        if fd >= 0:
+            os.close(fd)
+        temporary.unlink(missing_ok=True)
+    return True
+
+
+def main(argv: list[str]) -> int:
+    if len(argv) > 2:
+        print(f"usage: {Path(argv[0]).name} [settings.yml]", file=sys.stderr)
+        return 2
+
+    path = Path(argv[1]) if len(argv) == 2 else Path("/etc/searxng/settings.yml")
+    try:
+        changed = migrate_settings(path)
+    except (OSError, UnicodeError, ValueError) as exc:
+        print(f"SearXNG settings migration failed: {exc}", file=sys.stderr)
+        return 1
+
+    if changed:
+        print("Added use_default_settings inheritance to retained SearXNG settings")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main(sys.argv))

scripts/odysseus-webhook

@@ -2,7 +2,7 @@
 """odysseus-webhook — shell wrapper for scheduled-task webhook tokens.
 
 Tasks in the scheduled-task system can carry a `webhook_token`. Any
-HTTP POST to `/api/webhook/<token>` fires the task. This CLI lists,
+HTTP POST to `/api/tasks/<task-id>/webhook/<token>` fires the task. This CLI lists,
 rotates, and revokes those tokens.
 
     odysseus-webhook list                       # tasks that have a token
@@ -21,6 +21,7 @@ quiet_logs()
 
 import argparse, json, logging, os, secrets, sys
 from pathlib import Path
+from urllib.parse import quote
 
 try:
     from core.database import SessionLocal, ScheduledTask
@@ -53,6 +54,14 @@ def _summary(t: "ScheduledTask", reveal: bool = False) -> dict:
     }
 
 
+def _task_webhook_url(base: str, task_id: str, token: str) -> str:
+    """Build the live task-route URL without leaking ids into path syntax."""
+    root = (base or "http://localhost:7000").rstrip("/")
+    task_part = quote(str(task_id), safe="")
+    token_part = quote(str(token), safe="")
+    return f"{root}/api/tasks/{task_part}/webhook/{token_part}"
+
+
 def cmd_list(args):
     db = SessionLocal()
     try:
@@ -109,8 +118,7 @@ def cmd_url(args):
             fail(f"no task with id {args.id!r}")
         if not t.webhook_token:
             fail(f"task {args.id!r} has no webhook token (rotate one first)")
-        base = (args.base or "http://localhost:7000").rstrip("/")
-        url = f"{base}/api/webhook/{t.webhook_token}"
+        url = _task_webhook_url(args.base, t.id, t.webhook_token)
         emit({
             "task_id": t.id,
             "name": t.name,

services/docs/service.py

@@ -50,16 +50,46 @@ class DocsService:
             List of DocChunk objects
         """
         results = self.rag.search(query, k=top_k)
-        return [
-            DocChunk(
-                text=r.get("text", r.get("content", "")),
-                source=r.get("source", r.get("metadata", {}).get("source", "unknown")),
-                score=r.get("score", 0.0),
-                metadata=r.get("metadata"),
+        chunks = []
+
+        for result in results:
+            if not isinstance(result, dict):
+                continue
+
+            metadata = result.get("metadata")
+            if not isinstance(metadata, dict):
+                metadata = {}
+
+            text = result.get("document")
+            if text is None:
+                text = result.get("text")
+            if text is None:
+                text = result.get("content")
+            if text is None:
+                text = ""
+
+            source = result.get("source")
+            if source is None:
+                source = metadata.get("source")
+            if source is None:
+                source = "unknown"
+
+            score = result.get("similarity")
+            if score is None:
+                score = result.get("score")
+            if score is None:
+                score = 0.0
+
+            chunks.append(
+                DocChunk(
+                    text=text,
+                    source=source,
+                    score=score,
+                    metadata=metadata,
+                )
             )
-            for r in results
-            if isinstance(r, dict)
-        ]
+
+        return chunks
 
     async def index(self, directory: str) -> IndexResult:
         """
@@ -73,8 +103,8 @@ class DocsService:
         """
         result = self.rag.index_personal_documents(directory)
         return IndexResult(
-            indexed=result.get("indexed", 0),
-            failed=result.get("failed", 0),
+            indexed=result.get("indexed_count", result.get("indexed", 0)),
+            failed=result.get("failed_count", result.get("failed", 0)),
             errors=result.get("errors", []),
         )
 

services/memory/skill_importer.py

@@ -1,16 +1,18 @@
 """Import SKILL.md bundles from public GitHub (or skills.sh → GitHub) URLs."""
 from __future__ import annotations
 
+import ipaddress
 import logging
 import os
-import re
+import time
 from dataclasses import dataclass
-from typing import Dict, List, Optional, Tuple
+from typing import Dict, Iterable, List, Optional, Tuple, cast
 from urllib.parse import quote, urljoin, urlparse
 
+import httpcore
 import httpx
 
-from src.url_safety import check_outbound_url
+from src.url_safety import _default_resolver, check_outbound_url
 
 logger = logging.getLogger(__name__)
 
@@ -25,6 +27,7 @@ TEXT_NAMES = {"skill.md", "license", "license.md", "readme.md"}
 _GITHUB_HOSTS = frozenset({
     "github.com", "www.github.com", "api.github.com", "raw.githubusercontent.com",
 })
+_SKILLS_SH_HOSTS = frozenset({"skills.sh", "www.skills.sh"})
 
 
 def _github_host(url: str) -> str:
@@ -72,18 +75,158 @@ def _is_text_file(name: str) -> bool:
 _MAX_FETCH_REDIRECTS = 5
 
 
-def _check_fetch_url(url: str) -> None:
-    """SSRF guard for skill-import fetches (defense-in-depth).
+def _validated_ips(raw_ips: List[str]) -> List[ipaddress._BaseAddress]:
+    """Parse and de-duplicate one resolver snapshot in resolver order."""
+    ips: List[ipaddress._BaseAddress] = []
+    seen = set()
+    for raw in raw_ips:
+        if not isinstance(raw, str):
+            continue
+        try:
+            ip = ipaddress.ip_address(raw.split("%", 1)[0])
+        except ValueError:
+            continue
+        if ip in seen:
+            continue
+        seen.add(ip)
+        ips.append(ip)
+    return ips
 
-    Skill bundles only ever come from public GitHub, never an internal
-    address, so block private/loopback/link-local targets on every hop —
-    matching the hardened web-fetch path in
-    ``services/search/content.py:_get_public_url`` rather than the lenient
-    default used for admin-configured model endpoints.
-    """
-    ok, reason = check_outbound_url(url, block_private=True)
+
+def _resolve_and_check_url(url: str) -> List[ipaddress._BaseAddress]:
+    """Return the exact address snapshot approved for one fetch hop."""
+    resolved_ips: List[str] = []
+
+    def _recording_resolver(host: str) -> List[str]:
+        answers = list(_default_resolver(host))
+        resolved_ips[:] = answers
+        return answers
+
+    ok, reason = check_outbound_url(
+        url,
+        block_private=True,
+        resolver=_recording_resolver,
+    )
     if not ok:
-        raise SkillImportError(reason)
+        raise SkillImportError(f"outbound URL blocked: {reason}")
+
+    pinned_ips = _validated_ips(resolved_ips)
+    if not pinned_ips:
+        raise SkillImportError("outbound URL blocked: host did not resolve to a usable address")
+    return pinned_ips
+
+
+# Backward compatibility alias for tests importing _check_fetch_url directly
+_check_fetch_url = _resolve_and_check_url
+
+
+class _PinnedBackend(httpcore.NetworkBackend):
+    """Connect only to addresses from one validated DNS snapshot."""
+
+    def __init__(self, ips: List[ipaddress._BaseAddress]):
+        self._ips = [str(ip) for ip in ips]
+        self._real = httpcore.SyncBackend()
+
+    def connect_tcp(
+        self,
+        host: str,
+        port: int,
+        timeout: float | None = None,
+        local_address: str | None = None,
+        socket_options=None,
+    ):
+        deadline = None if timeout is None else time.monotonic() + timeout
+        last_exc: Optional[Exception] = None
+        for ip in self._ips:
+            remaining = None if deadline is None else max(0.0, deadline - time.monotonic())
+            try:
+                return self._real.connect_tcp(
+                    ip,
+                    port,
+                    remaining,
+                    local_address,
+                    socket_options,
+                )
+            except (httpcore.ConnectError, httpcore.ConnectTimeout) as exc:
+                last_exc = exc
+                if deadline is not None and time.monotonic() >= deadline:
+                    break
+        if last_exc is not None:
+            raise last_exc
+        raise httpcore.ConnectError("no validated address available")
+
+    def connect_unix_socket(self, path, timeout=None, socket_options=None):
+        return self._real.connect_unix_socket(path, timeout, socket_options)
+
+    def sleep(self, seconds: float) -> None:
+        return self._real.sleep(seconds)
+
+
+_HTTPCORE_TO_HTTPX_EXC = {
+    httpcore.ConnectError: httpx.ConnectError,
+    httpcore.ConnectTimeout: httpx.ConnectTimeout,
+    httpcore.LocalProtocolError: httpx.LocalProtocolError,
+    httpcore.NetworkError: httpx.NetworkError,
+    httpcore.PoolTimeout: httpx.PoolTimeout,
+    httpcore.ProtocolError: httpx.ProtocolError,
+    httpcore.ProxyError: httpx.ProxyError,
+    httpcore.ReadError: httpx.ReadError,
+    httpcore.ReadTimeout: httpx.ReadTimeout,
+    httpcore.RemoteProtocolError: httpx.RemoteProtocolError,
+    httpcore.TimeoutException: httpx.TimeoutException,
+    httpcore.UnsupportedProtocol: httpx.UnsupportedProtocol,
+    httpcore.WriteError: httpx.WriteError,
+    httpcore.WriteTimeout: httpx.WriteTimeout,
+}
+
+
+class _PinnedTransport(httpx.BaseTransport):
+    """Pin socket connects while preserving URL authority, Host, and TLS SNI."""
+
+    def __init__(self, ips: List[ipaddress._BaseAddress]):
+        self._pinned_ips = list(ips)
+        self._pool = httpcore.ConnectionPool(
+            ssl_context=httpx.create_ssl_context(),
+            http1=True,
+            http2=False,
+            network_backend=_PinnedBackend(ips),
+        )
+
+    def handle_request(self, request: httpx.Request) -> httpx.Response:
+        core_request = httpcore.Request(
+            method=request.method,
+            url=httpcore.URL(
+                scheme=request.url.raw_scheme,
+                host=request.url.raw_host,
+                port=request.url.port,
+                target=request.url.raw_path,
+            ),
+            headers=request.headers.raw,
+            content=request.stream,
+            extensions=request.extensions,
+        )
+        core_response = None
+        try:
+            core_response = self._pool.handle_request(core_request)
+            content = b"".join(cast(Iterable[bytes], core_response.stream))
+        except Exception as exc:
+            mapped = _HTTPCORE_TO_HTTPX_EXC.get(type(exc))
+            if mapped is not None:
+                raise mapped(str(exc)) from exc
+            raise
+        finally:
+            if core_response is not None:
+                core_response.close()
+
+        return httpx.Response(
+            status_code=core_response.status,
+            headers=core_response.headers,
+            content=content,
+            extensions=core_response.extensions,
+        )
+
+    def close(self) -> None:
+        self._pool.close()
 
 
 def _get_checked(
@@ -100,49 +243,76 @@ def _get_checked(
     hand lets us re-validate every hop, closing that blind-SSRF gap.
     """
     current = url
-    with httpx.Client(follow_redirects=False, timeout=timeout) as client:
-        for _ in range(_MAX_FETCH_REDIRECTS + 1):
-            _check_fetch_url(current)
+    for _ in range(_MAX_FETCH_REDIRECTS + 1):
+        pinned_ips = _resolve_and_check_url(current)
+        with httpx.Client(
+            transport=_PinnedTransport(pinned_ips),
+            follow_redirects=False,
+            timeout=timeout,
+        ) as client:
             r = client.get(current, headers=headers)
-            if r.status_code in (301, 302, 303, 307, 308):
-                location = r.headers.get("location")
-                if not location:
-                    return r
-                current = urljoin(str(r.url), location)
-                continue
-            return r
+
+        if r.status_code in (301, 302, 303, 307, 308):
+            location = r.headers.get("location")
+            if not location:
+                return r
+            current = urljoin(str(r.url), location)
+            continue
+        return r
     raise SkillImportError("too many redirects while fetching skill bundle")
 
 
 def parse_skill_source(url: str) -> ResolvedSource:
     """Normalize skills.sh / GitHub web URLs into owner/repo/ref/path."""
-    raw = (url or "").strip()
-    if not raw:
+    url = (url or "").strip()
+    if not url:
         raise SkillImportError("URL is required")
 
-    # skills.sh often links to GitHub; try to unwrap ?url= or redirect target later.
-    if "skills.sh" in raw and "github.com" not in raw:
-        r = _get_checked(raw, timeout=20.0)
+    # ``urlparse`` only reports an unambiguous scheme when the URL carries the
+    # ``scheme://`` form. Opaque schemes (``mailto:``, ``javascript:``) and a
+    # schemeless ``host:port`` both parse a "scheme" that is not one, so they
+    # fall through to the host check below and are rejected on the host instead.
+    scheme = urlparse(url).scheme.lower()
+    if scheme not in ("http", "https"):
+        if scheme and url.lower().startswith(f"{scheme}://"):
+            raise SkillImportError(f"unsupported URL scheme: {scheme}")
+        # Schemeless "github.com/owner/repo" — accept only a supported host.
+        rough_host = (urlparse("//" + url).hostname or "").lower()
+        if rough_host not in _GITHUB_HOSTS and rough_host not in _SKILLS_SH_HOSTS:
+            raise SkillImportError("Only GitHub or skills.sh URLs are supported")
+        url = "https://" + url
+
+    parsed = urlparse(url)
+    hostname = (parsed.hostname or "").lower()
+    if hostname not in _GITHUB_HOSTS and hostname not in _SKILLS_SH_HOSTS:
+        raise SkillImportError("Only GitHub or skills.sh URLs are supported")
+
+    # A skills.sh link is only usable if it redirects to an exact supported
+    # GitHub host. Scraping the page body for a github.com link cannot work:
+    # skill pages only ever link the repository root, never the skill's
+    # subdirectory, so the scrape resolves every skill in a repo to the same
+    # (wrong) bundle. Fail with an actionable message instead.
+    if hostname in _SKILLS_SH_HOSTS:
+        r = _get_checked(url, timeout=20.0)
         if r.status_code >= 400:
             raise _github_response_error(r)
         final = str(r.url)
-        _assert_github_url(final, context="redirect target")
-        # Page may embed a github link; prefer final URL if redirected.
-        if "github.com" in final:
-            raw = final
-        else:
-            m = re.search(r"https?://github\.com/[^\s\"')]+", r.text or "")
-            if m:
-                raw = m.group(0).rstrip(".,)")
+        if _github_host(final) not in _GITHUB_HOSTS:
+            raise SkillImportError(
+                "skills.sh did not redirect to GitHub — open the skill's "
+                "repository on GitHub, navigate to the exact skill folder or "
+                "SKILL.md file, and paste that URL; the repository-root link "
+                "alone is not sufficient"
+            )
+        url = final
 
-    parsed = urlparse(raw)
-    host = _github_host(raw)
-    if host not in _GITHUB_HOSTS:
-        raise SkillImportError(
-            "Only GitHub URLs are supported (https://github.com/... or raw.githubusercontent.com/...)"
-        )
+    # Update parsed and hostname to reflect the new GitHub URL
+    parsed = urlparse(url)
+    hostname = (parsed.hostname or "").lower()
 
-    if host == "raw.githubusercontent.com":
+    _assert_github_url(url)
+
+    if hostname == "raw.githubusercontent.com":
         # /owner/repo/ref/path/to/file
         bits = [p for p in parsed.path.split("/") if p]
         if len(bits) < 4:

services/search/content.py

@@ -2,22 +2,18 @@
 
 import copy
 import io
-import ipaddress
 import json
 import os
 import re
 import logging
-import socket
-import ssl
 from datetime import datetime, timedelta
-from typing import Iterable, List, cast
-from urllib.parse import urljoin, urlparse
+from typing import List
 
 import httpx
-import httpcore
 from bs4 import BeautifulSoup
 
 from src.constants import WEB_FETCH_SOFT_MAX_BYTES, WEB_FETCH_HARD_MAX_BYTES, WEB_FETCH_USER_AGENT
+from src import outbound_fetch as _outbound_fetch
 
 from .analytics import RateLimitError, error_logger
 from .cache import (
@@ -29,336 +25,40 @@ from .cache import (
 
 logger = logging.getLogger(__name__)
 
-_PRIVATE_NETWORKS = (
-    ipaddress.ip_network("0.0.0.0/8"),
-    ipaddress.ip_network("10.0.0.0/8"),
-    ipaddress.ip_network("127.0.0.0/8"),
-    ipaddress.ip_network("169.254.0.0/16"),
-    ipaddress.ip_network("172.16.0.0/12"),
-    ipaddress.ip_network("192.168.0.0/16"),
-    ipaddress.ip_network("::1/128"),
-    ipaddress.ip_network("fc00::/7"),
-    ipaddress.ip_network("fe80::/10"),
-)
+def _is_private_address(addr):
+    return _outbound_fetch._is_private_address(addr)
 
 
-def _is_private_address(addr: ipaddress._BaseAddress) -> bool:
-    if isinstance(addr, ipaddress.IPv6Address) and addr.ipv4_mapped is not None:
-        addr = addr.ipv4_mapped
-    return (
-        addr.is_private
-        or addr.is_loopback
-        or addr.is_link_local
-        or addr.is_reserved
-        or addr.is_multicast
-        or addr.is_unspecified
-        or any(addr in net for net in _PRIVATE_NETWORKS)
+def _resolve_hostname_ips(hostname):
+    return _outbound_fetch._resolve_hostname_ips(hostname)
+
+
+def _public_http_url(url):
+    return _outbound_fetch._public_http_url(url, resolver=_resolve_hostname_ips)
+
+
+def _resolve_public_ips(url):
+    return _outbound_fetch._resolve_public_ips(url, resolver=_resolve_hostname_ips)
+
+
+_PinnedBackend = _outbound_fetch._PinnedBackend
+_PinnedTransport = _outbound_fetch._PinnedTransport
+BodyTooLargeError = _outbound_fetch.BodyTooLargeError
+_CappedFetch = _outbound_fetch._CappedFetch
+
+
+def _get_public_url(url, headers, timeout, max_redirects=5, max_bytes=None):
+    return _outbound_fetch._get_public_url(
+        url,
+        headers=headers,
+        timeout=timeout,
+        max_redirects=max_redirects,
+        max_bytes=max_bytes,
+        resolve_public_ips=_resolve_public_ips,
+        transport_factory=_PinnedTransport,
     )
 
 
-def _resolve_hostname_ips(hostname: str) -> list[ipaddress._BaseAddress]:
-    try:
-        infos = socket.getaddrinfo(hostname, None)
-    except Exception:
-        return []
-    out = []
-    for info in infos:
-        try:
-            out.append(ipaddress.ip_address(info[4][0]))
-        except Exception:
-            continue
-    return out
-
-
-def _public_http_url(url: str) -> bool:
-    try:
-        parsed = urlparse(url)
-        if parsed.scheme not in ("http", "https"):
-            return False
-        host = (parsed.hostname or "").strip()
-        if not host:
-            return False
-        lower = host.lower()
-        if lower in ("localhost", "metadata", "metadata.google.internal"):
-            return False
-        if lower.endswith((".local", ".localhost", ".internal", ".lan", ".intranet")):
-            return False
-        try:
-            return not _is_private_address(ipaddress.ip_address(host))
-        except ValueError:
-            pass
-        addrs = _resolve_hostname_ips(host)
-        return bool(addrs) and not any(_is_private_address(a) for a in addrs)
-    except Exception:
-        return False
-
-
-def _resolve_public_ips(url: str) -> list[ipaddress._BaseAddress]:
-    parsed = urlparse(url)
-    if parsed.scheme not in ("http", "https") or not parsed.hostname:
-        raise httpx.RequestError(f"Blocked non-public URL: {url}")
-    host = (parsed.hostname or "").strip().lower()
-    if host in ("localhost", "metadata", "metadata.google.internal"):
-        raise httpx.RequestError(f"Blocked non-public hostname: {host}")
-    try:
-        ip = ipaddress.ip_address(host)
-        if _is_private_address(ip):
-            raise httpx.RequestError(f"Blocked non-public IP literal: {host}")
-        return [ip]
-    except httpx.RequestError:
-        raise
-    except ValueError:
-        pass
-    addrs = _resolve_hostname_ips(host)
-    if not addrs or any(_is_private_address(a) for a in addrs):
-        raise httpx.RequestError(f"Blocked non-public URL: {url}")
-    return addrs
-
-
-class _PinnedBackend(httpcore.NetworkBackend):
-    """Network backend that connects to a pre-resolved IP.
-
-    httpcore derives the TLS SNI and the ``Host`` header from the URL's
-    origin, not from the host argument passed to ``connect_tcp``. So
-    routing the TCP connect to a resolved IP while leaving the URL
-    untouched keeps SNI / vhost behaviour correct and closes the
-    DNS-rebinding TOCTOU between the SSRF check and the connect.
-    """
-
-    def __init__(self, ip: ipaddress._BaseAddress):
-        self._ip = str(ip)
-        self._real = httpcore.SyncBackend()
-
-    def connect_tcp(
-        self,
-        host: str,
-        port: int,
-        timeout: float | None = None,
-        local_address: str | None = None,
-        socket_options=None,
-    ):
-        return self._real.connect_tcp(
-            self._ip, port, timeout, local_address, socket_options
-        )
-
-    def connect_unix_socket(self, path, timeout=None, socket_options=None):
-        return self._real.connect_unix_socket(path, timeout, socket_options)
-
-    def sleep(self, seconds: float) -> None:
-        return self._real.sleep(seconds)
-
-
-# Map httpcore exception classes to their httpx equivalents. Built
-# once at import time from the public exception classes; avoids any
-# import of httpx's private transport machinery. httpcore's
-# ``ConnectionNotAvailable`` is a pool-internal signal (the pool will
-# close and retry on its own) — we never expect to see it surface to
-# a transport caller, so it has no httpx counterpart here.
-_HTTPCORE_TO_HTTPX_EXC = {
-    httpcore.ConnectError: httpx.ConnectError,
-    httpcore.ConnectTimeout: httpx.ConnectTimeout,
-    httpcore.LocalProtocolError: httpx.LocalProtocolError,
-    httpcore.NetworkError: httpx.NetworkError,
-    httpcore.PoolTimeout: httpx.PoolTimeout,
-    httpcore.ProtocolError: httpx.ProtocolError,
-    httpcore.ProxyError: httpx.ProxyError,
-    httpcore.ReadError: httpx.ReadError,
-    httpcore.ReadTimeout: httpx.ReadTimeout,
-    httpcore.RemoteProtocolError: httpx.RemoteProtocolError,
-    httpcore.TimeoutException: httpx.TimeoutException,
-    httpcore.UnsupportedProtocol: httpx.UnsupportedProtocol,
-    httpcore.WriteError: httpx.WriteError,
-    httpcore.WriteTimeout: httpx.WriteTimeout,
-}
-
-
-class _PinnedTransport(httpx.BaseTransport):
-    """Transport that pins every TCP connect to a pre-resolved IP.
-
-    Uses only the public ``httpcore`` and ``httpx`` APIs — no
-    subclassing of ``httpx.HTTPTransport``, no reads of private
-    ``httpcore.ConnectionPool`` attributes, no imports from
-    ``httpx private transport internals``. The URL is passed through unchanged so SNI
-    / vhost work as if httpx had been given the hostname directly;
-    only the TCP destination is pinned, closing the DNS-rebinding
-    TOCTOU between the SSRF check and the connect.
-    """
-
-    def __init__(self, ip: ipaddress._BaseAddress, *, http2: bool = False):
-        self._pool = httpcore.ConnectionPool(
-            ssl_context=ssl.create_default_context(),
-            http1=True,
-            http2=http2,
-            network_backend=_PinnedBackend(ip),
-        )
-
-    def __enter__(self):
-        self._pool.__enter__()
-        return self
-
-    def __exit__(self, exc_type=None, exc_value=None, traceback=None) -> None:
-        self._pool.__exit__(exc_type, exc_value, traceback)
-
-    def handle_request(self, request: httpx.Request) -> httpx.Response:
-        httpcore_req = httpcore.Request(
-            method=request.method,
-            url=httpcore.URL(
-                scheme=request.url.raw_scheme,
-                host=request.url.raw_host,
-                port=request.url.port,
-                target=request.url.raw_path,
-            ),
-            headers=request.headers.raw,
-            content=request.stream,
-            extensions=request.extensions,
-        )
-        try:
-            httpcore_resp = self._pool.handle_request(httpcore_req)
-            # Eager materialisation matches the original
-            # ``response.text`` usage in fetch_webpage_content. The
-            # sync pool's stream is a plain Iterable[bytes] despite
-            # the httpcore type hint unioning the async variant.
-            content = b"".join(cast(Iterable[bytes], httpcore_resp.stream))
-        except Exception as exc:
-            mapped = _HTTPCORE_TO_HTTPX_EXC.get(type(exc))
-            if mapped is not None:
-                raise mapped(str(exc)) from exc
-            raise
-
-        return httpx.Response(
-            status_code=httpcore_resp.status,
-            headers=httpcore_resp.headers,
-            content=content,
-            extensions=httpcore_resp.extensions,
-        )
-
-    def close(self) -> None:
-        self._pool.close()
-
-class BodyTooLargeError(Exception):
-    """The server declared a body larger than the hard fetch ceiling."""
-
-    def __init__(self, url: str, declared_bytes: int):
-        self.url = url
-        self.declared_bytes = declared_bytes
-        super().__init__(
-            f"response body is {declared_bytes:,} bytes, over the "
-            f"{WEB_FETCH_HARD_MAX_BYTES:,}-byte hard cap"
-        )
-
-
-class _CappedFetch:
-    """Result of a size-capped streaming GET.
-
-    Carries just what fetch_webpage_content needs from an httpx.Response,
-    plus the cap bookkeeping: the (possibly truncated) body, whether the
-    cap cut it short, and the size the server declared via Content-Length
-    (wire bytes; None when absent).
-    """
-
-    __slots__ = ("status_code", "headers", "content", "truncated",
-                 "declared_bytes", "encoding", "url")
-
-    def __init__(self, status_code, headers, content, truncated,
-                 declared_bytes, encoding, url):
-        self.status_code = status_code
-        self.headers = headers
-        self.content = content
-        self.truncated = truncated
-        self.declared_bytes = declared_bytes
-        self.encoding = encoding
-        self.url = url
-
-    @property
-    def text(self) -> str:
-        return self.content.decode(self.encoding or "utf-8", errors="replace")
-
-    def raise_for_status(self):
-        if self.status_code >= 400:
-            request = httpx.Request("GET", self.url)
-            raise httpx.HTTPStatusError(
-                f"HTTP {self.status_code} for {self.url}",
-                request=request,
-                response=httpx.Response(self.status_code, request=request),
-            )
-
-
-def _get_public_url(url: str, headers: dict, timeout: int, max_redirects: int = 5,
-                    max_bytes: int = None) -> "_CappedFetch":
-    """Capped streaming GET with SSRF-guarded, DNS-pinned manual redirects.
-
-    Each hop is resolved once, validated as public, and then the actual TCP
-    connection is pinned to that resolved IP. The request URL is left unchanged
-    so Host and TLS SNI keep the original hostname.
-    """
-    cap = min(max_bytes or WEB_FETCH_SOFT_MAX_BYTES, WEB_FETCH_HARD_MAX_BYTES)
-    current = url
-    for _ in range(max_redirects + 1):
-        ips = _resolve_public_ips(current)
-
-        # Force identity transfer-encoding. With gzip/deflate the wire bytes
-        # and Content-Length can be a small fraction of the decoded body, so a
-        # tiny compressed response could pass the hard-cap preflight and then
-        # expand past the ceiling in one decoded chunk before the streamed cap
-        # below can slice it.
-        req_headers = dict(headers or {})
-        req_headers["Accept-Encoding"] = "identity"
-
-        with httpx.Client(
-            headers=req_headers,
-            timeout=timeout,
-            follow_redirects=False,
-            transport=_PinnedTransport(ips[0]),
-        ) as client:
-            with client.stream("GET", current) as response:
-                if response.status_code in (301, 302, 303, 307, 308):
-                    location = response.headers.get("location")
-                    if not location:
-                        return _CappedFetch(response.status_code, response.headers, b"",
-                                            False, None, response.encoding, str(response.url))
-                    current = urljoin(str(response.url), location)
-                    continue
-
-                # A server can ignore the identity request and still return a
-                # compressed body; httpx.iter_bytes would then decode it, and a
-                # tiny gzip can balloon into one decoded chunk far past the cap.
-                # Refuse compressed Content-Encoding so the streamed cap stays
-                # a real memory bound.
-                enc = (response.headers.get("content-encoding") or "").strip().lower()
-                if enc and enc != "identity":
-                    raise httpx.RequestError(
-                        f"Refusing compressed response (Content-Encoding: {enc}) after "
-                        "requesting identity: cannot bound decoded body size",
-                        request=httpx.Request("GET", current),
-                    )
-
-                declared = None
-                raw_len = response.headers.get("content-length")
-                if raw_len and raw_len.isdigit():
-                    declared = int(raw_len)
-
-                if declared is not None and declared > WEB_FETCH_HARD_MAX_BYTES:
-                    raise BodyTooLargeError(current, declared)
-
-                chunks = []
-                read = 0
-                truncated = False
-                for chunk in response.iter_bytes():
-                    read += len(chunk)
-                    if read > cap:
-                        keep = cap - (read - len(chunk))
-                        if keep > 0:
-                            chunks.append(chunk[:keep])
-                        truncated = True
-                        break
-                    chunks.append(chunk)
-
-                return _CappedFetch(response.status_code, response.headers,
-                                    b"".join(chunks), truncated, declared,
-                                    response.encoding, str(response.url))
-
-    raise httpx.RequestError("Too many redirects", request=httpx.Request("GET", current))
-
 # PDF extraction (optional dependency)
 try:
     from pdfminer.high_level import extract_text as pdf_extract_text

src/agent_runs.py

@@ -17,13 +17,14 @@ close / navigation / refresh). It does NOT survive a server restart.
 import asyncio
 import json
 import logging
+import uuid
 from typing import AsyncGenerator, Dict, Optional
 
 logger = logging.getLogger(__name__)
 
 
 class _Run:
-    __slots__ = ("buffer", "subscribers", "status", "task", "evict_task")
+    __slots__ = ("buffer", "subscribers", "status", "task", "evict_task", "run_id")
 
     def __init__(self) -> None:
         self.buffer: list = []          # ordered SSE event strings (replay log)
@@ -31,6 +32,9 @@ class _Run:
         self.status: str = "running"    # running | done | error | stopped
         self.task: Optional[asyncio.Task] = None
         self.evict_task: Optional[asyncio.Task] = None
+        # Stable across every subscription/replay of this exact detached run.
+        # The browser uses it to make local cost accounting replay-idempotent.
+        self.run_id: str = uuid.uuid4().hex
 
 
 _RUNS: Dict[str, _Run] = {}
@@ -53,13 +57,24 @@ def _publish(run: _Run, ev: str) -> None:
             pass
 
 
-def _schedule_evict(session_id: str) -> None:
+def _wake_run_subscribers(run: _Run) -> None:
+    """Close subscribers even when the drain task never reached its body."""
+    for q in list(run.subscribers):
+        try:
+            q.put_nowait((None, None))
+        except Exception:
+            pass
+
+
+def _schedule_evict(session_id: str, expected_run: Optional[_Run] = None) -> None:
     """(Re)arm a grace-period eviction for a terminal run with no subscribers.
     Identity-checked so a run that gets replaced/reused is never evicted by a
     stale timer."""
     run = _RUNS.get(session_id)
     if run is None:
         return
+    if expected_run is not None and run is not expected_run:
+        return
     if run.evict_task and not run.evict_task.done():
         run.evict_task.cancel()
 
@@ -85,25 +100,38 @@ def get_status(session_id: str) -> Optional[str]:
     return r.status if r else None
 
 
-async def _drain(session_id: str, agen: AsyncGenerator[str, None],
+def get_run_id(session_id: str) -> Optional[str]:
+    """Return the opaque identity of the current detached run, if present."""
+    r = _RUNS.get(session_id)
+    return r.run_id if r else None
+
+
+def get_active_run(session_id: str) -> Optional[_Run]:
+    """Return the exact active run currently registered for a session."""
+    r = _RUNS.get(session_id)
+    return r if r and r.status == "running" else None
+
+
+async def _drain(session_id: str, run: _Run, agen: AsyncGenerator[str, None],
                  prev_task: Optional[asyncio.Task] = None) -> None:
     """Pull every event from the wrapped generator into the run buffer, fanning
     each out to live subscribers. Runs to completion regardless of subscribers."""
-    run = _RUNS.get(session_id)
-    if run is None:
-        return
+    subscribers_woken = False
+
+    def _wake_subscribers() -> None:
+        nonlocal subscribers_woken
+        if subscribers_woken:
+            return
+        subscribers_woken = True
+        _wake_run_subscribers(run)
+
     # If this run replaced an in-flight one (rapid double-send), wait for that
     # one to fully finish first. Its CancelledError handler calls aclose(), which
     # persists its partial response — letting it complete before we start writing
     # keeps the two runs' session saves sequential instead of interleaved.
-    if prev_task is not None and not prev_task.done():
-        try:
-            await asyncio.wait({prev_task})
-        except asyncio.CancelledError:
-            raise            # our own cancellation — propagate
-        except Exception:
-            pass
     try:
+        if prev_task is not None and not prev_task.done():
+            await asyncio.wait({prev_task})
         async for ev in agen:
             _publish(run, ev)
         if run.status == "running":
@@ -116,6 +144,16 @@ async def _drain(session_id: str, agen: AsyncGenerator[str, None],
             await agen.aclose()
         except Exception:
             pass
+        # A rapid third replacement can cancel this task while it is still
+        # waiting for its predecessor. Close this run's subscribers promptly,
+        # but keep the task alive until the predecessor finishes so the next
+        # run still observes the transitive session-save ordering barrier.
+        _wake_subscribers()
+        if prev_task is not None and not prev_task.done():
+            try:
+                await asyncio.shield(prev_task)
+            except (asyncio.CancelledError, Exception):
+                pass
     except Exception as e:
         logger.error("[agent-run] %s failed: %s", session_id, e, exc_info=True)
         run.status = "error"
@@ -127,15 +165,11 @@ async def _drain(session_id: str, agen: AsyncGenerator[str, None],
         _publish(run, "data: [DONE]\n\n")
     finally:
         # Wake every subscriber with the end sentinel so their SSE closes.
-        for q in list(run.subscribers):
-            try:
-                q.put_nowait((None, None))
-            except Exception:
-                pass
+        _wake_subscribers()
         # Run is terminal — arm the grace timer so it (and its buffer) is
         # eventually freed even if nobody ever reconnects. subscribe() cancels
         # this on connect and re-arms on disconnect.
-        _schedule_evict(session_id)
+        _schedule_evict(session_id, run)
 
 
 def start(session_id: str, agen: AsyncGenerator[str, None]) -> _Run:
@@ -145,20 +179,37 @@ def start(session_id: str, agen: AsyncGenerator[str, None]) -> _Run:
     prev_task: Optional[asyncio.Task] = None
     if prev:
         if prev.task and not prev.task.done():
+            # A task cancelled before its first instruction never enters
+            # _drain(), so its except/finally blocks cannot update status or
+            # wake a response already bound to this exact run. Terminalize it
+            # synchronously before cancelling; _drain's cleanup is idempotent
+            # when the task had already started.
+            if prev.status == "running":
+                prev.status = "stopped"
+                _wake_run_subscribers(prev)
             prev.task.cancel()
             prev_task = prev.task   # new run awaits this before it starts writing
         if prev.evict_task and not prev.evict_task.done():
             prev.evict_task.cancel()
     run = _Run()
     _RUNS[session_id] = run
-    run.task = asyncio.create_task(_drain(session_id, agen, prev_task))
+    run.task = asyncio.create_task(_drain(session_id, run, agen, prev_task))
     return run
 
 
-async def subscribe(session_id: str) -> AsyncGenerator[str, None]:
+async def subscribe(
+    session_id: str,
+    expected_run: Optional[_Run] = None,
+) -> AsyncGenerator[str, None]:
     """Replay the run's buffer from the start, then stream live until it ends.
-    Safe to call repeatedly (reconnect) and from multiple clients at once."""
-    run = _RUNS.get(session_id)
+    Safe to call repeatedly (reconnect) and from multiple clients at once.
+
+    ``expected_run`` binds a lazy StreamingResponse body to the same run whose
+    identity was put in its response headers. Without that binding, a rapid
+    replacement between response construction and body iteration could replay
+    the replacement run under the prior run's identity.
+    """
+    run = expected_run or _RUNS.get(session_id)
     if run is None:
         return
     q: asyncio.Queue = asyncio.Queue()
@@ -201,12 +252,19 @@ async def subscribe(session_id: str) -> AsyncGenerator[str, None]:
         # Last subscriber gone on a finished run — (re)arm eviction so the
         # buffer doesn't linger indefinitely.
         if not run.subscribers and run.status != "running":
-            _schedule_evict(session_id)
+            _schedule_evict(session_id, run)
 
 
-def stop(session_id: str) -> bool:
-    """Cancel an in-flight run (the wrapped generator saves its partial)."""
+def stop(session_id: str, expected_run_id: Optional[str] = None) -> bool:
+    """Cancel the matching in-flight run (which saves its partial output).
+
+    A stale browser may issue Stop after another tab has replaced the session's
+    run. Once the caller knows its opaque run identity, fail closed rather than
+    cancelling that newer run.
+    """
     run = _RUNS.get(session_id)
+    if not expected_run_id or run is None or run.run_id != expected_run_id:
+        return False
     if run and run.task and not run.task.done():
         run.task.cancel()
         return True

src/agent_tools/admin_tools.py

@@ -510,7 +510,12 @@ async def do_manage_settings(content: str, owner: Optional[str] = None) -> Dict:
         # set/get/list/delete operate on the REAL app settings (the same store
         # the Settings panel writes), so changing a model / voice / search
         # engine / reminder channel from chat actually takes effect.
-        from src.settings import load_settings, save_settings, DEFAULT_SETTINGS
+        from src.settings import (
+            DEFAULT_SETTINGS,
+            RETIRED_SETTING_KEYS,
+            load_settings,
+            save_settings,
+        )
 
         # Secrets/credentials the agent must NOT write: kept read-only (masked)
         # so API keys never flow through chat. User sets these in the panel.
@@ -562,6 +567,9 @@ async def do_manage_settings(content: str, owner: Optional[str] = None) -> Dict:
                 return k2
             return _ALIASES_SET.get(k2, (k or "").strip())
 
+        def _is_managed_key(key):
+            return key in DEFAULT_SETTINGS and key not in RETIRED_SETTING_KEYS
+
         _ENUMS = {
             "image_quality": ["low", "medium", "high"],
             "reminder_channel": ["browser", "email", "ntfy", "webhook"],
@@ -624,14 +632,18 @@ async def do_manage_settings(content: str, owner: Optional[str] = None) -> Dict:
 
         if action == "list":
             s = load_settings()
-            shown = {k: _mask(k, v) for k, v in s.items() if k in DEFAULT_SETTINGS and not isinstance(v, dict)}
+            shown = {
+                k: _mask(k, v)
+                for k, v in s.items()
+                if _is_managed_key(k) and not isinstance(v, dict)
+            }
             return {"response": f"{len(shown)} settings (use get/set with a key)", "settings": shown, "exit_code": 0}
 
         elif action == "get":
             key = _resolve(args.get("key", ""))
             if not key:
                 return {"error": "key is required", "exit_code": 1}
-            if key not in DEFAULT_SETTINGS:
+            if not _is_managed_key(key):
                 return {"error": f"Unknown setting '{args.get('key')}'. Use action='list' to see them.", "exit_code": 1}
             val = load_settings().get(key, DEFAULT_SETTINGS.get(key))
             return {"response": f"{key} = {_mask(key, val)}", "value": _mask(key, val), "exit_code": 0}
@@ -642,11 +654,11 @@ async def do_manage_settings(content: str, owner: Optional[str] = None) -> Dict:
             if not raw:
                 return {"error": "key is required", "exit_code": 1}
             key = _resolve(raw)
-            if key not in DEFAULT_SETTINGS:
+            if not _is_managed_key(key):
                 return {"error": f"Unknown setting '{raw}'. Use action='list' to see available settings.", "exit_code": 1}
             if _is_secret(key):
                 return {"response": f"'{key}' is a credential/secret. For security I can't set it from chat. Open Settings and set it there.", "exit_code": 0}
-            # Structured settings (dicts/lists like keybinds, default_model_fallbacks)
+            # Structured settings (dicts/lists like keybinds or vision fallbacks)
             # have no safe scalar coercion; _coerce would pass a bare string
             # straight through and clobber the structure. Refuse them here; they're
             # edited in their dedicated panels. (reset/delete still restore the
@@ -675,7 +687,7 @@ async def do_manage_settings(content: str, owner: Optional[str] = None) -> Dict:
 
         elif action == "delete" or action == "reset":
             key = _resolve(args.get("key", ""))
-            if key not in DEFAULT_SETTINGS:
+            if not _is_managed_key(key):
                 return {"error": f"Unknown setting '{args.get('key')}'.", "exit_code": 1}
             if _is_secret(key):
                 return {"response": f"'{key}' is a credential. Reset it in the panel.", "exit_code": 0}

src/agent_tools/document_tools.py

@@ -2,6 +2,7 @@ from typing import Any, Dict, List, Optional
 import logging
 import re
 from src.constants import MAX_READ_CHARS
+from src.tool_approvals import document_content_digest
 from src.tool_utils import _parse_tool_args, get_upload_handler
 from src.upload_handler import reserve_upload_references
 
@@ -80,6 +81,40 @@ def _most_recent_owned_document(db, Document, owner: Optional[str], active_only:
     return q.order_by(Document.updated_at.desc()).first()
 
 
+def _approved_document_version_error(doc: Any, ctx: dict) -> Optional[Dict]:
+    """Reject a sealed document action when its target changed meanwhile."""
+    expected_version = ctx.get("expected_document_version")
+    expected_digest = (
+        str(ctx.get("expected_document_digest") or "").strip().lower()
+    )
+    if expected_version is None and not expected_digest:
+        return None
+    try:
+        version_unchanged = (
+            expected_version is None
+            or int(getattr(doc, "version_count", -1)) == int(expected_version)
+        )
+    except (TypeError, ValueError):
+        version_unchanged = False
+    content_unchanged = True
+    if expected_digest:
+        content_unchanged = (
+            doc is not None
+            and document_content_digest(getattr(doc, "current_content", ""))
+            == expected_digest
+        )
+    if version_unchanged and content_unchanged:
+        return None
+    return {
+        "error": (
+            "The target document changed after this action was proposed. "
+            "Review the latest version and request the edit again."
+        ),
+        "exit_code": 1,
+        "document_changed": True,
+    }
+
+
 # ---------------------------------------------------------------------------
 # Document tools — create/update/edit/suggest living documents
 # ---------------------------------------------------------------------------
@@ -454,6 +489,12 @@ class UpdateDocumentTool:
             doc = None
             if target_id:
                 doc = _get_owned_document(db, Document, target_id, owner)
+            if (
+                not doc
+                and target_id
+                and ctx.get("expected_document_version") is not None
+            ):
+                return _approved_document_version_error(None, ctx)
             if not doc:
                 doc = _most_recent_owned_document(db, Document, owner)
                 if doc:
@@ -463,6 +504,10 @@ class UpdateDocumentTool:
             if not doc:
                 return {"error": "No documents exist to update"}
 
+            version_error = _approved_document_version_error(doc, ctx)
+            if version_error:
+                return version_error
+
             is_email_doc = doc.language == "email" or _looks_like_email_document(doc.current_content or "", doc.title or "")
             new_content = _coerce_email_document_content(doc.current_content or "", content) if is_email_doc else content.strip()
             if is_email_doc:
@@ -530,6 +575,12 @@ class EditDocumentTool:
             doc = None
             if target_id:
                 doc = _get_owned_document(db, Document, target_id, owner)
+            if (
+                not doc
+                and target_id
+                and ctx.get("expected_document_version") is not None
+            ):
+                return _approved_document_version_error(None, ctx)
             if not doc:
                 # Fallback: most recently updated document. Avoids "no active doc" errors
                 # after server restart or when the agent loses track of which doc to edit.
@@ -541,6 +592,10 @@ class EditDocumentTool:
             if not doc:
                 return {"error": "No documents exist to edit"}
 
+            version_error = _approved_document_version_error(doc, ctx)
+            if version_error:
+                return version_error
+
             is_email_doc = doc.language == "email" or _looks_like_email_document(doc.current_content or "", doc.title or "")
             blank_find_edits = [e for e in edits if not (e.get("find") or "").strip()]
             if blank_find_edits:
@@ -677,6 +732,10 @@ class SuggestDocumentTool:
             if not doc:
                 return {"error": f"Document {target_id} not found"}
 
+            version_error = _approved_document_version_error(doc, ctx)
+            if version_error:
+                return version_error
+
             # Validate that FIND text exists in document
             valid = []
             for s in suggestions:

src/agent_tools/model_interaction_tools.py

@@ -64,7 +64,10 @@ async def chat_with_model(content: str, session_id: Optional[str] = None, owner:
         return {"model": model, "response": response}
     except Exception as e:
         logger.error(f"chat_with_model failed: {e}")
-        return {"error": f"Failed to get response from {model_spec}: {e}"}
+        return {
+            "error": f"Failed to get response from {model_spec}: {e}",
+            "untrusted_content": True,
+        }
 
 
 async def ask_teacher(content: str, session_id: Optional[str] = None, owner: Optional[str] = None) -> Dict:
@@ -110,7 +113,10 @@ async def ask_teacher(content: str, session_id: Optional[str] = None, owner: Opt
         return {"model": model, "response": response, "teacher": True}
     except Exception as e:
         logger.error(f"ask_teacher failed: {e}")
-        return {"error": f"Teacher call failed ({model_spec}): {e}"}
+        return {
+            "error": f"Teacher call failed ({model_spec}): {e}",
+            "untrusted_content": True,
+        }
 
 
 async def list_models(content: str, session_id: Optional[str] = None, owner: Optional[str] = None) -> Dict:

src/agent_tools/session_tools.py

@@ -240,7 +240,10 @@ async def send_to_session(content: str, session_id: Optional[str] = None, owner:
         }
     except Exception as e:
         logger.error(f"send_to_session failed: {e}")
-        return {"error": f"Failed to send to session: {e}"}
+        return {
+            "error": f"Failed to send to session: {e}",
+            "untrusted_content": True,
+        }
 
 async def manage_session(content: str, session_id: Optional[str] = None, owner: Optional[str] = None) -> Dict:
     """Manage sessions: rename, archive, delete, important, truncate, fork.

src/agent_tools/web_tools.py

@@ -66,6 +66,7 @@ class WebSearchTool:
             return {
                 "error": f"web_search failed: {type(e).__name__}: {str(e) or 'no details'}",
                 "exit_code": 1,
+                "untrusted_content": True,
             }
         if progress_cb:
             await progress_cb({
@@ -136,7 +137,11 @@ class WebFetchTool:
 
         if not text:
             if err:
-                return {"error": f"web_fetch: {url}: {err}", "exit_code": 1}
+                return {
+                    "error": f"web_fetch: {url}: {err}",
+                    "exit_code": 1,
+                    "untrusted_content": True,
+                }
             return {"error": f"web_fetch: {url}: no readable text content (not HTML, or the page needs JS/login)", "exit_code": 1}
 
         # Tell the model when the download budget cut the body short and how

src/ai_interaction.py

@@ -324,7 +324,10 @@ async def do_pipeline(content: str, session_id: Optional[str] = None, owner: Opt
         }
     except Exception as e:
         logger.error(f"pipeline failed at step {len(step_outputs) + 1}: {e}")
-        return {"error": f"Pipeline failed at step {len(step_outputs) + 1}: {e}"}
+        return {
+            "error": f"Pipeline failed at step {len(step_outputs) + 1}: {e}",
+            "untrusted_content": True,
+        }
 
 
 # ---------------------------------------------------------------------------
@@ -1089,7 +1092,10 @@ async def do_generate_image(content: str, session_id: Optional[str] = None, owne
                     error_text = err_json.get("error", {}).get("message", error_text) if isinstance(err_json.get("error"), dict) else str(err_json.get("error", error_text))
                 except Exception:
                     pass
-                return {"error": f"Image generation failed ({resp.status_code}): {error_text}"}
+                return {
+                    "error": f"Image generation failed ({resp.status_code}): {error_text}",
+                    "untrusted_content": True,
+                }
 
             data = resp.json()
             images = data.get("data", [])
@@ -1173,7 +1179,10 @@ async def do_generate_image(content: str, session_id: Optional[str] = None, owne
     except httpx.TimeoutException:
         return {"error": "Image generation timed out (300s). The model may be overloaded — try again or use quality=low."}
     except Exception as e:
-        return {"error": f"Image generation error: {str(e)}"}
+        return {
+            "error": f"Image generation error: {str(e)}",
+            "untrusted_content": True,
+        }
 
 
 async def do_edit_image(
@@ -1310,7 +1319,10 @@ async def do_edit_image(
                     error_text = err_json.get("detail") or err_json.get("error") or error_text
                 except Exception:
                     pass
-                return {"error": f"Image edit fallback failed ({fallback_resp.status_code}): {error_text}"}
+                return {
+                    "error": f"Image edit fallback failed ({fallback_resp.status_code}): {error_text}",
+                    "untrusted_content": True,
+                }
             fallback_data = fallback_resp.json()
             image_b64 = fallback_data.get("image")
             if not image_b64:
@@ -1394,7 +1406,10 @@ async def do_edit_image(
                                 "model for attached-image prompts."
                             )
                         }
-                return {"error": f"Image edit failed ({resp.status_code}): {error_text}"}
+                return {
+                    "error": f"Image edit failed ({resp.status_code}): {error_text}",
+                    "untrusted_content": True,
+                }
 
             data = resp.json()
             images = data.get("data", [])
@@ -1434,7 +1449,10 @@ async def do_edit_image(
     except httpx.TimeoutException:
         return {"error": "Image edit timed out. The model may still be loading or overloaded."}
     except Exception as e:
-        return {"error": f"Image edit error: {str(e)}"}
+        return {
+            "error": f"Image edit error: {str(e)}",
+            "untrusted_content": True,
+        }
 
 
 # ---------------------------------------------------------------------------

src/auth_helpers.py

@@ -4,6 +4,8 @@ import os
 from typing import Optional
 from fastapi import Request, HTTPException
 
+from src.owner_identity import auth_disabled, effective_storage_owner
+
 
 def get_current_user(request: Request) -> Optional[str]:
     """Get current username from request state (set by auth middleware)."""
@@ -56,7 +58,17 @@ def _auth_disabled() -> bool:
     """True when the operator has explicitly turned off auth via .env.
     Mirrors the AUTH_ENABLED parse in app.py / core/middleware.py so the
     three call sites agree on what "off" means."""
-    return os.getenv("AUTH_ENABLED", "true").lower() == "false"
+    return auth_disabled()
+
+
+def storage_owner_for_request(request: Request) -> Optional[str]:
+    """Resolve the storage owner for code paths that need an owner bucket.
+
+    This does not replace route authentication. It only gives auth-disabled
+    no-login mode a stable storage identity instead of writing new data as
+    legacy NULL/ownerless state.
+    """
+    return effective_storage_owner(effective_user(request))
 
 
 def require_user(request: Request) -> str:

src/bg_monitor.py

@@ -15,6 +15,7 @@ import json
 import logging
 
 from src import bg_jobs
+from src.prompt_security import untrusted_context_message
 
 logger = logging.getLogger(__name__)
 
@@ -25,6 +26,16 @@ POLL_INTERVAL_S = 5
 _FOLLOWUP_MAX_ROUNDS = 12
 
 
+def _background_result_message(rec):
+    inject = (
+        f"[Background job {rec['id']} finished]\n\n"
+        f"{bg_jobs.result_text(rec)}\n\n"
+        "Continue the task using this output. Don't repeat work that's already done. "
+        "If the task is now complete, give the user the final result."
+    )
+    return untrusted_context_message("background job output", inject)
+
+
 async def _drain_agent(sess, messages):
     """Run the agent loop headless against a session. Returns
     (final_prose, tool_events) — tool_events in the same shape the live chat
@@ -62,13 +73,19 @@ async def _drain_agent(sess, messages):
             round_num = d.get("round", round_num)
         elif d.get("type") == "tool_output":
             # Mirror the live chat's tool_event shape (chat_routes / chatRenderer).
-            tool_events.append({
+            tool_event = {
                 "round": round_num,
                 "tool": d.get("tool"),
                 "command": d.get("command"),
                 "output": d.get("output"),
                 "exit_code": d.get("exit_code"),
-            })
+            }
+            if isinstance(d.get("ask_user"), dict):
+                # Preserve exact-approval cards from a tainted background-job
+                # continuation so the user can authorize the sealed action on
+                # the next foreground turn instead of losing it headlessly.
+                tool_event["ask_user"] = d["ask_user"]
+            tool_events.append(tool_event)
     return full, tool_events
 
 
@@ -101,14 +118,8 @@ async def _run_followup(rec: dict) -> bool:
     except Exception:
         pass
 
-    inject = (
-        f"[Background job {rec['id']} finished]\n\n"
-        f"{bg_jobs.result_text(rec)}\n\n"
-        "Continue the task using this output. Don't repeat work that's already done. "
-        "If the task is now complete, give the user the final result."
-    )
     context = sess.get_context_messages()
-    context.append({"role": "user", "content": inject})
+    context.append(_background_result_message(rec))
 
     full, tool_events = await _drain_agent(sess, context)
 

src/builtin_actions.py

@@ -810,13 +810,27 @@ async def action_tidy_research(owner: str, **kwargs) -> Tuple[str, bool]:
 
     Research history lives entirely in data/deep_research/<id>.json and is NOT
     backed by chat-session rows — so a file must never be deleted just because
-    no chat session matches its id. Only prune files that fail to load."""
+    no chat session matches its id. Only prune files that fail to load.
+
+    A broken file has no readable owner stamp, so it cannot be matched against
+    `owner`. Clearing one is privileged: admins and the single-user operator
+    (AUTH_ENABLED=false) may, a regular user may not, and neither may anyone
+    during the pre-setup window before an admin exists.
+    """
     try:
         from pathlib import Path
         import json as _json
+        from src.tool_security import owner_is_admin_or_single_user
         research_dir = Path(DEEP_RESEARCH_DIR)
         if not research_dir.exists():
             raise TaskNoop("no research directory")
+        if not owner_is_admin_or_single_user(owner):
+            # Return before the glob rather than filtering inside the loop: the
+            # loop reports "none broken" off an empty `removed`, which reaches
+            # Activity as a false report to a user whose files it skipped, and a
+            # regular user need not read every owner's file to learn it may
+            # delete none of them.
+            raise TaskNoop("not permitted to remove unattributable research files")
         files = list(research_dir.glob("*.json"))
         removed = []
         for p in files:

src/chat_processor.py

@@ -381,7 +381,10 @@ class ChatProcessor:
                         )
                         if len(rag_content) > 10000:
                             rag_content = rag_content[:10000] + "\n[Truncated]"
-                        preface.append(untrusted_context_message("retrieved documents", rag_content))
+                        preface.append(untrusted_context_message(
+                            "retrieved documents",
+                            rag_content,
+                        ))
             except Exception as e:
                 logger.warning(f"RAG retrieval failed: {e}")
 
@@ -459,12 +462,38 @@ class ChatProcessor:
         skip_url_fetch = len(message) > 2000 or len(non_yt_urls) > 3
         if not skip_url_fetch:
             for url in non_yt_urls:
-                result = fetch_webpage_content(url)
+                try:
+                    result = fetch_webpage_content(url)
+                except Exception:
+                    # The URL and exception can both contain signed-query
+                    # credentials or response-controlled text. Keep the log
+                    # diagnostic stable as well as the model-facing context.
+                    logger.warning("Automatic URL fetch failed while building context")
+                    result = {"success": False, "error": ""}
                 if result.get('success'):
                     content = result.get('content', '')[:10000]
                     preface.append(untrusted_context_message(
                         f"web page: {url}",
                         f"Content from {url}:\n\n{content}",
+                        provenance_origin="external",
+                    ))
+                else:
+                    # A failed automatic URL fetch is context too. Never pass
+                    # exception text or response-controlled diagnostics back to
+                    # the model: reduce the result to a small transport-owned
+                    # status and explicitly state that the page was not read.
+                    error = str(result.get("error") or "")
+                    status = "the page was unavailable"
+                    status_match = re.match(r"^HTTP\s+(\d{3})\b", error)
+                    if status_match:
+                        status = f"the server returned HTTP {status_match.group(1)}"
+                    elif error.startswith("TooLarge:"):
+                        status = "the response exceeded the fetch size limit"
+                    elif error.startswith("Rate limit"):
+                        status = "the request was rate limited"
+                    preface.append(untrusted_context_message(
+                        "web page fetch failure",
+                        f"A linked page was not read: {status}.",
                     ))
 
         # Skills index — progressive disclosure. Only injected when the
@@ -488,6 +517,9 @@ class ChatProcessor:
                     for s in sorted(by_cat[cat], key=lambda x: x["name"]):
                         desc = s.get("description") or ""
                         lines.append(f"    - {s['name']}: {desc}" if desc else f"    - {s['name']}")
-                preface.append(untrusted_context_message("available skills index", "\n".join(lines)))
+                preface.append(untrusted_context_message(
+                    "available skills index",
+                    "\n".join(lines),
+                ))
 
         return preface, rag_sources, web_sources

src/context_compactor.py

@@ -282,7 +282,9 @@ def trim_for_context(messages: List[Dict], context_length: int, reserve_tokens:
     if essential_system:
         sys_text = essential_system[0].get("content", "")
         if len(sys_text) > 2000:
-            essential_system[0] = {"role": "system", "content": sys_text[:2000] + "\n[System prompt truncated for context limits]"}
+            truncated_system = dict(essential_system[0])
+            truncated_system["content"] = sys_text[:2000] + "\n[System prompt truncated for context limits]"
+            essential_system[0] = truncated_system
             trimmed = essential_system + convo_msgs
             if estimate_tokens(trimmed) <= budget:
                 return _sanitize_tool_messages(essential_system + protected_msgs + convo_msgs)
@@ -325,6 +327,9 @@ async def maybe_compact(
     messages: List[Dict],
     headers: Optional[Dict] = None,
     owner: Optional[str] = None,
+    *,
+    persist: bool = True,
+    compaction_state: Optional[Dict[str, Any]] = None,
 ) -> tuple:
     """Check context usage and compact if above threshold.
 
@@ -416,7 +421,17 @@ async def maybe_compact(
     # offset — session.history INCLUDES the system messages, but
     # split_point is indexed against convo_msgs which does NOT. Without
     # this, the slice drops the leading system message(s).
-    _update_session_history(session, split_point, summary, system_msg_count=len(system_msgs))
+    if compaction_state is not None:
+        compaction_state.update({
+            "split_point": split_point,
+            "summary": summary,
+            "system_msg_count": len(system_msgs),
+            "applied": False,
+        })
+    if persist:
+        _update_session_history(session, split_point, summary, system_msg_count=len(system_msgs))
+        if compaction_state is not None:
+            compaction_state["applied"] = True
 
     new_used = estimate_tokens(compacted)
     logger.info(
@@ -427,6 +442,51 @@ async def maybe_compact(
     return compacted, context_length, True
 
 
+def apply_compaction_state(session, compaction_state: Optional[Dict[str, Any]]) -> bool:
+    """Persist a route-specific compaction after that route commits output.
+
+    Candidate prompts may be compacted speculatively while an explicit
+    foreground fallback chain is being tried.  Persisting at construction time
+    would let an unavailable route rewrite history before another route answers,
+    so callers hold this small plan and apply only the winning route's plan.
+    """
+
+    state = compaction_state if isinstance(compaction_state, dict) else None
+    if not state or state.get("applied"):
+        return False
+    summary = state.get("summary")
+    split_point = state.get("split_point")
+    system_msg_count = state.get("system_msg_count", 0)
+    if not isinstance(summary, str) or not isinstance(split_point, int):
+        return False
+    _update_session_history(
+        session,
+        split_point,
+        summary,
+        system_msg_count=system_msg_count if isinstance(system_msg_count, int) else 0,
+    )
+    state["applied"] = True
+    return True
+
+
+def apply_compaction_state_for_session(
+    session_id: Optional[str],
+    compaction_state: Optional[Dict[str, Any]],
+) -> bool:
+    """Resolve an in-memory session and apply a deferred compaction plan."""
+
+    if not session_id:
+        return False
+    try:
+        from core.models import get_session_manager_instance
+
+        manager = get_session_manager_instance()
+        session = manager.get_session(session_id) if manager else None
+    except Exception:
+        session = None
+    return apply_compaction_state(session, compaction_state) if session else False
+
+
 def _update_session_history(session, split_point: int, summary: str,
                             system_msg_count: int = 0):
     """Update the in-memory session history after compaction.

src/endpoint_resolver.py

@@ -5,6 +5,7 @@ Consolidates the 4+ copies of normalize_base / resolve_endpoint logic into one p
 """
 
 import json
+import ipaddress
 import logging
 import socket
 import subprocess
@@ -27,6 +28,43 @@ _NON_CHAT_MODEL = (
 )
 
 
+def endpoint_cost_tracked(url: str, endpoint_kind: Optional[str] = None) -> bool:
+    """Return whether token cost should be tracked for a concrete route.
+
+    This is intentionally a non-secret route classification.  It mirrors the
+    frontend's local/subscription exclusions without exposing endpoint URLs to
+    message metadata.
+    """
+
+    try:
+        parsed = urlparse(url or "")
+        host = (parsed.hostname or "").lower().rstrip(".")
+        path = (parsed.path or "").rstrip("/")
+    except Exception:
+        return False
+    if not host:
+        return False
+    if host == "chatgpt.com" and (
+        path == "/backend-api/codex" or path.startswith("/backend-api/codex/")
+    ):
+        return False
+    kind = str(endpoint_kind or "auto").strip().lower()
+    if kind == "local":
+        return False
+    if kind in {"api", "proxy"}:
+        return True
+    if host in {"localhost", "0.0.0.0", "host.docker.internal"} or host.endswith(".local"):
+        return False
+    try:
+        ip = ipaddress.ip_address(host)
+        return ip.is_global
+    except ValueError:
+        pass
+    if "." not in host:
+        return False
+    return True
+
+
 def _first_chat_model(models) -> Optional[str]:
     """First model that isn't an embedding/tts/etc.; falls back to models[0]."""
     for m in (models or []):
@@ -396,10 +434,14 @@ def resolve_endpoint(
         db.close()
 
 
-def resolve_endpoint_by_id(
-    ep_id: str, model: Optional[str] = None, owner: Optional[str] = None
-) -> Optional[Tuple[str, str, Dict]]:
-    """Resolve a specific endpoint id (+ optional model) to (chat_url, model, headers).
+def _resolve_endpoint_by_id_with_descriptor(
+    ep_id: str,
+    model: Optional[str] = None,
+    owner: Optional[str] = None,
+    *,
+    require_exact_model: bool = False,
+) -> Optional[Tuple[Tuple[str, str, Dict], dict]]:
+    """Resolve a concrete endpoint/model plus its non-secret descriptor.
 
     Returns None if the endpoint doesn't exist or is disabled. Used to turn
     a configured fallback entry ({endpoint_id, model}) into a dispatch target.
@@ -426,15 +468,34 @@ def resolve_endpoint_by_id(
         chat_url = build_chat_url(base)
         headers = build_headers(api_key, base)
         m = (model or "").strip()
-        # Drop a model the user disabled on the endpoint, then pick the first
-        # enabled chat model rather than a hidden one.
-        if m and m in _endpoint_hidden_models(ep):
-            m = ""
-        if not m:
-            m = _first_chat_model(_endpoint_enabled_models(ep)) or ""
+        enabled_models = _endpoint_enabled_models(ep)
+        if require_exact_model:
+            # Explicit foreground fallback entries are concrete choices. A
+            # hidden or known-missing model must disable the entry instead of
+            # silently substituting another model from the endpoint.
+            if not m or m in _endpoint_hidden_models(ep):
+                return None
+            if enabled_models and m not in enabled_models:
+                return None
+        else:
+            # Legacy Utility/Vision chains retain their model-repair behavior.
+            if m and m in _endpoint_hidden_models(ep):
+                m = ""
+            if not m:
+                m = _first_chat_model(enabled_models) or ""
         if not m:
             return None
-        return chat_url, m, headers
+        return (
+            (chat_url, m, headers),
+            {
+                "endpoint_id": ep.id,
+                "endpoint_label": getattr(ep, "name", None) or ep.id,
+                "endpoint_cost_tracked": endpoint_cost_tracked(
+                    chat_url,
+                    getattr(ep, "endpoint_kind", None),
+                ),
+            },
+        )
     except Exception as e:
         logger.debug(f"Could not resolve endpoint {ep_id}: {e}")
         return None
@@ -442,11 +503,101 @@ def resolve_endpoint_by_id(
         db.close()
 
 
-def resolve_chat_fallback_candidates(owner: Optional[str] = None) -> list:
-    """Compatibility shim for the retired default-chat fallback chain."""
+def resolve_endpoint_by_id(
+    ep_id: str,
+    model: Optional[str] = None,
+    owner: Optional[str] = None,
+    *,
+    require_exact_model: bool = False,
+) -> Optional[Tuple[str, str, Dict]]:
+    """Resolve a specific endpoint id (+ optional model) to its runtime route."""
 
-    del owner
-    return []
+    resolved = _resolve_endpoint_by_id_with_descriptor(
+        ep_id,
+        model,
+        owner=owner,
+        require_exact_model=require_exact_model,
+    )
+    return resolved[0] if resolved else None
+
+
+def resolve_route_descriptor(
+    endpoint_url: str,
+    model: str,
+    headers: Optional[Dict] = None,
+    owner: Optional[str] = None,
+) -> dict:
+    """Return the visible endpoint identity for an already-resolved route.
+
+    Headers are compared only inside the process so two endpoints using the
+    same provider URL/model but different credentials remain distinguishable.
+    No credential material is returned or logged.
+    """
+
+    if not endpoint_url or not model:
+        return {
+            "endpoint_id": None,
+            "endpoint_label": "Selected route",
+            "endpoint_cost_tracked": endpoint_cost_tracked(endpoint_url),
+        }
+    db = SessionLocal()
+    try:
+        q = db.query(ModelEndpoint).filter(ModelEndpoint.is_enabled == True)
+        if owner:
+            from src.auth_helpers import owner_filter
+            q = owner_filter(q, ModelEndpoint, owner)
+        expected = (endpoint_url.rstrip("/"), model, headers or {})
+        for ep in q.all():
+            resolved = _resolve_endpoint_by_id_with_descriptor(
+                ep.id,
+                model,
+                owner=owner,
+                require_exact_model=True,
+            )
+            if not resolved:
+                continue
+            candidate, descriptor = resolved
+            actual = (candidate[0].rstrip("/"), candidate[1], candidate[2] or {})
+            if actual == expected:
+                return descriptor
+    except Exception as e:
+        logger.debug("Could not identify selected endpoint route: %s", e)
+    finally:
+        db.close()
+    return {
+        "endpoint_id": None,
+        "endpoint_label": "Selected route",
+        "endpoint_cost_tracked": endpoint_cost_tracked(endpoint_url),
+    }
+
+
+def resolve_route_descriptor_by_id(
+    endpoint_id: str,
+    endpoint_url: str,
+    model: str,
+    headers: Optional[Dict] = None,
+    owner: Optional[str] = None,
+) -> Optional[dict]:
+    """Resolve a selected route's identity without relying on row order.
+
+    The explicit endpoint id is still verified against the resolved runtime
+    route. This prevents stale or mismatched request metadata from being used
+    for attribution while disambiguating endpoints whose routes are otherwise
+    identical.
+    """
+
+    resolved = _resolve_endpoint_by_id_with_descriptor(
+        endpoint_id,
+        model,
+        owner=owner,
+        require_exact_model=True,
+    )
+    if not resolved:
+        return None
+    candidate, descriptor = resolved
+    expected = ((endpoint_url or "").rstrip("/"), model, headers or {})
+    actual = (candidate[0].rstrip("/"), candidate[1], candidate[2] or {})
+    return descriptor if actual == expected else None
 
 
 def resolve_utility_fallback_candidates(owner: Optional[str] = None) -> list:
@@ -460,17 +611,62 @@ def resolve_vision_fallback_candidates(owner: Optional[str] = None) -> list:
 
 
 def _resolve_fallback_candidates(setting_key: str, owner: Optional[str] = None) -> list:
-    out = []
     try:
         from src.settings import get_user_setting, load_settings
         settings = load_settings()
         chain = get_user_setting(setting_key, owner or "", settings.get(setting_key) or []) or []
     except Exception:
-        return out
-    for entry in chain:
+        return []
+    return resolve_fallback_entries(chain, owner=owner)
+
+
+def resolve_fallback_entries(
+    entries,
+    owner: Optional[str] = None,
+    *,
+    require_exact_model: bool = False,
+) -> list:
+    """Resolve ordered endpoint/model entries within the caller's owner scope."""
+
+    out = []
+    for entry in entries or []:
         if not isinstance(entry, dict):
             continue
-        resolved = resolve_endpoint_by_id(entry.get("endpoint_id", ""), entry.get("model", ""), owner=owner)
-        if resolved:
+        resolved = resolve_endpoint_by_id(
+            entry.get("endpoint_id", ""),
+            entry.get("model", ""),
+            owner=owner,
+            require_exact_model=require_exact_model,
+        )
+        if resolved and resolved not in out:
             out.append(resolved)
     return out
+
+
+def resolve_fallback_entries_with_descriptors(
+    entries,
+    owner: Optional[str] = None,
+    *,
+    require_exact_model: bool = False,
+) -> list:
+    """Resolve ordered entries while retaining safe endpoint provenance."""
+
+    out = []
+    seen = []
+    for entry in entries or []:
+        if not isinstance(entry, dict):
+            continue
+        resolved = _resolve_endpoint_by_id_with_descriptor(
+            entry.get("endpoint_id", ""),
+            entry.get("model", ""),
+            owner=owner,
+            require_exact_model=require_exact_model,
+        )
+        if not resolved:
+            continue
+        candidate, descriptor = resolved
+        if any(candidate == prior for prior in seen):
+            continue
+        seen.append(candidate)
+        out.append((candidate, descriptor))
+    return out

src/foreground_model_routing.py

@@ -1,22 +1,155 @@
-"""Foreground Chat and Agent model-routing policy.
+"""Explicit foreground Chat and Agent model-routing policy."""
 
-The selected session model is strict by default. Historical
-``default_model_fallbacks`` values remain stored for compatibility, but this
-policy intentionally does not read or migrate them.
-"""
+from dataclasses import dataclass
+from typing import Any, Collection, Dict, FrozenSet, Optional, Tuple
 
-from typing import Any, Dict, Optional
+from src.endpoint_resolver import (
+    endpoint_cost_tracked,
+    resolve_fallback_entries,
+    resolve_fallback_entries_with_descriptors,
+    resolve_route_descriptor,
+    resolve_route_descriptor_by_id,
+)
+
+_DEFAULT_FALLBACK_ENTRY_RESOLVER = resolve_fallback_entries
+
+
+FOREGROUND_FALLBACK_ENABLED_KEY = "foreground_fallback_enabled"
+FOREGROUND_FALLBACK_LIST_KEY = "foreground_model_fallbacks"
+FOREGROUND_AVAILABILITY_STATUSES: FrozenSet[int] = frozenset({
+    408, 425, 429, 500, 502, 503, 504, 507, 508, 529,
+})
+MAX_FOREGROUND_FALLBACKS = 10
+
+
+@dataclass(frozen=True)
+class ForegroundModelPolicy:
+    """Resolved per-user foreground fallback policy."""
+
+    enabled: bool = False
+    fallback_candidates: Tuple[tuple, ...] = ()
+    fallback_descriptors: Tuple[dict, ...] = ()
+    eligible_statuses: FrozenSet[int] = FOREGROUND_AVAILABILITY_STATUSES
+    fallback_on_empty: bool = False
+
+
+def _load_policy_preferences(owner: Optional[str]) -> dict:
+    """Load only preferences that explicitly belong to ``owner``.
+
+    The generic preferences loader intentionally treats a legacy flat store as
+    the single-user preferences object.  That compatibility must not cross an
+    authentication transition: once a named owner is present, foreground
+    fallback consent exists only in an actual ``_users[owner]`` dictionary.
+    """
+
+    from routes import prefs_routes
+
+    if owner is None:
+        prefs = prefs_routes._load_for_user(None)
+        return dict(prefs) if isinstance(prefs, dict) else {}
+
+    raw = prefs_routes._load()
+    users = raw.get("_users") if isinstance(raw, dict) else None
+    if not isinstance(users, dict):
+        return {}
+    prefs = users.get(owner)
+    return dict(prefs) if isinstance(prefs, dict) else {}
+
+
+def resolve_foreground_model_policy(
+    owner: Optional[str] = None,
+    allowed_models: Optional[Collection[str]] = None,
+) -> ForegroundModelPolicy:
+    """Resolve an explicit owner-scoped policy, failing closed to strict mode.
+
+    The policy is stored in user preferences even when authentication is
+    disabled. Historical ``default_model_fallbacks`` values are deliberately
+    unrelated and are never read or migrated.
+    """
+
+    try:
+        prefs = _load_policy_preferences(owner)
+    except Exception:
+        return ForegroundModelPolicy()
+
+    if prefs.get(FOREGROUND_FALLBACK_ENABLED_KEY) is not True:
+        return ForegroundModelPolicy()
+
+    entries = prefs.get(FOREGROUND_FALLBACK_LIST_KEY)
+    if not isinstance(entries, list) or not entries:
+        return ForegroundModelPolicy()
+    if allowed_models is not None:
+        allowed = frozenset(allowed_models)
+        entries = [
+            entry for entry in entries
+            if (
+                isinstance(entry, dict)
+                and isinstance(entry.get("model"), str)
+                and entry.get("model") in allowed
+            )
+        ]
+        if not entries:
+            return ForegroundModelPolicy()
+    entries = entries[:MAX_FOREGROUND_FALLBACKS]
+
+    if resolve_fallback_entries is not _DEFAULT_FALLBACK_ENTRY_RESOLVER:
+        # Preserve the long-standing resolver seam used by downstream tests and
+        # integrations. Production uses the descriptor-aware resolver below.
+        compatibility_candidates = resolve_fallback_entries(
+            entries,
+            owner=owner,
+            require_exact_model=True,
+        )
+        # Known limitation of this test-only seam: alignment matches on model
+        # alone, so when two entries share a model and the resolver skips the
+        # first, the surviving candidate inherits the skipped entry's
+        # endpoint_id. Production uses the descriptor-aware branch below,
+        # which is unaffected.
+        resolved_routes = []
+        remaining_entries = list(entries)
+        for candidate in compatibility_candidates:
+            matching_index = next(
+                (
+                    index for index, entry in enumerate(remaining_entries)
+                    if isinstance(entry, dict)
+                    and entry.get("model") == candidate[1]
+                ),
+                None,
+            )
+            matching_entry = (
+                remaining_entries.pop(matching_index)
+                if matching_index is not None
+                else {}
+            )
+            descriptor = {
+                "endpoint_id": matching_entry.get("endpoint_id"),
+                "endpoint_label": matching_entry.get("endpoint_id") or "Fallback route",
+                "endpoint_cost_tracked": endpoint_cost_tracked(candidate[0]),
+            }
+            resolved_routes.append((candidate, descriptor))
+    else:
+        resolved_routes = resolve_fallback_entries_with_descriptors(
+            entries,
+            owner=owner,
+            require_exact_model=True,
+        )
+    candidates = [candidate for candidate, _descriptor in resolved_routes]
+    if not candidates:
+        return ForegroundModelPolicy()
+
+    return ForegroundModelPolicy(
+        enabled=True,
+        fallback_candidates=tuple(candidates),
+        fallback_descriptors=tuple(
+            dict(descriptor) for _candidate, descriptor in resolved_routes
+        ),
+    )
 
 
 def resolve_foreground_fallback_candidates(owner: Optional[str] = None) -> list:
-    """Return fallback candidates for a foreground Chat or Agent request.
+    """Return only candidates explicitly enabled by the current user."""
 
-    Foreground routing is strict, so no alternate endpoint/model is eligible.
-    ``owner`` is accepted to keep this policy boundary owner-aware.
-    """
-
-    del owner
-    return []
+    return list(resolve_foreground_model_policy(owner).fallback_candidates)
 
 
 def build_foreground_model_candidates(
@@ -24,8 +157,50 @@ def build_foreground_model_candidates(
     model: str,
     headers: Optional[Dict[str, Any]] = None,
     owner: Optional[str] = None,
+    policy: Optional[ForegroundModelPolicy] = None,
 ) -> list:
     """Build the ordered candidate list for a foreground request."""
 
+    policy = policy or resolve_foreground_model_policy(owner)
     primary = (endpoint_url, model, headers or {})
-    return [primary] + resolve_foreground_fallback_candidates(owner=owner)
+    candidates = [primary]
+    for candidate in policy.fallback_candidates:
+        if candidate not in candidates:
+            candidates.append(candidate)
+    return candidates
+
+
+def build_foreground_route_descriptors(
+    endpoint_url: str,
+    model: str,
+    headers: Optional[Dict[str, Any]] = None,
+    owner: Optional[str] = None,
+    policy: Optional[ForegroundModelPolicy] = None,
+    selected_endpoint_id: Optional[str] = None,
+) -> list:
+    """Build safe route metadata parallel to foreground candidates."""
+
+    policy = policy or resolve_foreground_model_policy(owner)
+    selected = None
+    if selected_endpoint_id:
+        selected = resolve_route_descriptor_by_id(
+            selected_endpoint_id,
+            endpoint_url,
+            model,
+            headers or {},
+            owner=owner,
+        )
+    if selected is None:
+        selected = resolve_route_descriptor(endpoint_url, model, headers or {}, owner=owner)
+    primary = (endpoint_url, model, headers or {})
+    candidates = [primary]
+    descriptors = [selected]
+    for candidate, descriptor in zip(
+        policy.fallback_candidates,
+        policy.fallback_descriptors,
+    ):
+        if candidate in candidates:
+            continue
+        candidates.append(candidate)
+        descriptors.append(dict(descriptor))
+    return descriptors

src/integrations.py

@@ -719,7 +719,14 @@ async def execute_api_call(
         output = f"HTTP {status}\n{formatted}"
 
         if status >= 400:
-            return {"error": output, "exit_code": 1}
+            return {
+                "error": output,
+                "exit_code": 1,
+                # The error string includes the remote response body.  Preserve
+                # it for diagnostics, but make its provenance explicit so the
+                # agent gate does not treat HTTP failure as content-free.
+                "untrusted_content": True,
+            }
 
         return {"output": output, "exit_code": 0}
 

src/interactive_gate.py

@@ -63,8 +63,11 @@ _PASSIVE_EXACT_PATHS = {
     "/api/activity/heartbeat",
     "/api/client-perf",
     "/api/tasks/notifications",
+    "/api/tasks/runs/recent",
     "/api/research/active",
     "/api/email/urgency-state",
+    # UI idle poll sibling of urgency-state; must not pre-empt background tasks.
+    "/api/email/unread-state",
 }
 
 _PASSIVE_PREFIXES = (
@@ -74,6 +77,19 @@ _PASSIVE_PREFIXES = (
 )
 
 
+async def maybe_stop_background_tasks_for_heartbeat(stop_background) -> bool:
+    """Stop background work for browser activity only when the gate is enabled.
+
+    ``stop_background`` is injected by the application boundary so this policy
+    remains independently testable without importing the full FastAPI app.
+    """
+    if not _enabled():
+        return False
+
+    await stop_background(reason="browser heartbeat")
+    return True
+
+
 def should_track_interactive_request(path: str, method: str = "GET") -> bool:
     if not _enabled():
         return False

src/mcp_manager.py

@@ -530,6 +530,8 @@ class McpManager:
             "stderr": output if is_error else "",
             "exit_code": 1 if is_error else 0,
         }
+        if is_error and output:
+            result_dict["untrusted_content"] = True
         if images:
             result_dict["images"] = images
         return result_dict

src/mcp_oauth.py

@@ -15,18 +15,32 @@ from urllib.parse import urlparse, parse_qs
 
 logger = logging.getLogger(__name__)
 
+
+def _resolve_redirect_base() -> str:
+    """Origin the browser is sent back to after authorizing.
+
+    Falls back to the port the app binds natively (APP_PORT, read the same way
+    by app.py and launcher.py) rather than a fixed 7000: the macOS launcher
+    defaults to 7860, and a callback on the wrong port reaches nothing. The
+    hostname stays `localhost` rather than internal_api_base()'s 127.0.0.1 —
+    this URI is registered with the authorization server (via DCR, or by hand
+    for Google clients), so changing the host invalidates registrations that
+    already exist.
+    """
+    return (
+        os.environ.get("OAUTH_REDIRECT_BASE_URL")
+        or os.environ.get("APP_PUBLIC_URL")
+        or f"http://localhost:{os.environ.get('APP_PORT', '7000')}"
+    ).rstrip("/")
+
+
 # OAuth redirect URI registered with every authorization server via DCR. Loopback
 # is allowed for native/desktop clients (RFC 8252); remote users finish via the
-# paste-back flow. Deployments not reachable at http://localhost:7000 (custom
-# port, reverse proxy, or public domain) must set OAUTH_REDIRECT_BASE_URL (or
-# APP_PUBLIC_URL) to their externally reachable origin so the redirect lands back
-# on Odysseus. APP_PORT is intentionally not used: it is only the Docker host
-# port-map; the app always listens on 7000 inside the container.
-_REDIRECT_BASE = (
-    os.environ.get("OAUTH_REDIRECT_BASE_URL")
-    or os.environ.get("APP_PUBLIC_URL")
-    or "http://localhost:7000"
-).rstrip("/")
+# paste-back flow. Deployments whose externally reachable origin differs from the
+# port Odysseus binds — reverse proxy, public domain, or Docker, whose host port
+# map is invisible inside the container — must set OAUTH_REDIRECT_BASE_URL (or
+# APP_PUBLIC_URL), otherwise the redirect never lands back on Odysseus.
+_REDIRECT_BASE = _resolve_redirect_base()
 REDIRECT_URI = f"{_REDIRECT_BASE}/api/mcp/oauth/callback"
 
 # How long the background connect waits for the user to authorize before giving up.

src/model_capability_readers/base.py

@@ -290,17 +290,22 @@ def detect_vendor(base_url: Any = "", endpoint_kind: Any = "") -> str:
         return kind_map[kind]
 
     parsed = urlparse(compact_str(base_url))
-    host = (parsed.hostname or "").lower()
+    host = (parsed.hostname or "").lower().rstrip(".")
     port = parsed.port
-    if host.endswith("openrouter.ai"):
+
+    def host_matches(domain: str) -> bool:
+        domain = domain.lower().rstrip(".")
+        return host == domain or host.endswith(f".{domain}")
+
+    if host_matches("openrouter.ai"):
         return VENDOR_OPENROUTER
-    if host.endswith("openai.com"):
+    if host_matches("openai.com"):
         return VENDOR_OPENAI
-    if host.endswith("anthropic.com"):
+    if host_matches("anthropic.com"):
         return VENDOR_ANTHROPIC
-    if host.endswith("googleapis.com"):
+    if host_matches("googleapis.com"):
         return VENDOR_GOOGLE
-    if host.endswith("ollama.com") or port == 11434:
+    if host_matches("ollama.com") or port == 11434:
         return VENDOR_OLLAMA
     if port == 1234:
         return VENDOR_LMSTUDIO

src/outbound_fetch.py

@@ -0,0 +1,354 @@
+"""SSRF-guarded synchronous HTTP fetching primitives.
+
+This module owns outbound URL classification, one-resolution-per-hop DNS
+pinning, redirects, and response-body budgets.  It deliberately has no search
+or content-extraction dependencies so callers outside search can reuse the
+same transport boundary.
+"""
+
+from __future__ import annotations
+
+import ipaddress
+import socket
+import ssl
+from typing import Callable, Iterable, cast
+from urllib.parse import urljoin, urlparse
+
+import httpcore
+import httpx
+
+from src.constants import WEB_FETCH_HARD_MAX_BYTES, WEB_FETCH_SOFT_MAX_BYTES
+
+
+_PRIVATE_NETWORKS = (
+    ipaddress.ip_network("0.0.0.0/8"),
+    ipaddress.ip_network("10.0.0.0/8"),
+    ipaddress.ip_network("127.0.0.0/8"),
+    ipaddress.ip_network("169.254.0.0/16"),
+    ipaddress.ip_network("172.16.0.0/12"),
+    ipaddress.ip_network("192.168.0.0/16"),
+    ipaddress.ip_network("::1/128"),
+    ipaddress.ip_network("fc00::/7"),
+    ipaddress.ip_network("fe80::/10"),
+)
+
+
+def _is_private_address(addr: ipaddress._BaseAddress) -> bool:
+    if isinstance(addr, ipaddress.IPv6Address) and addr.ipv4_mapped is not None:
+        addr = addr.ipv4_mapped
+    return (
+        addr.is_private
+        or addr.is_loopback
+        or addr.is_link_local
+        or addr.is_reserved
+        or addr.is_multicast
+        or addr.is_unspecified
+        or any(addr in net for net in _PRIVATE_NETWORKS)
+    )
+
+
+def _resolve_hostname_ips(hostname: str) -> list[ipaddress._BaseAddress]:
+    try:
+        infos = socket.getaddrinfo(hostname, None)
+    except Exception:
+        return []
+    out = []
+    for info in infos:
+        try:
+            out.append(ipaddress.ip_address(info[4][0]))
+        except Exception:
+            continue
+    return out
+
+
+def _public_http_url(
+    url: str,
+    *,
+    resolver: Callable[[str], list[ipaddress._BaseAddress]] | None = None,
+) -> bool:
+    resolver = resolver or _resolve_hostname_ips
+    try:
+        parsed = urlparse(url)
+        if parsed.scheme not in ("http", "https"):
+            return False
+        host = (parsed.hostname or "").strip()
+        if not host:
+            return False
+        lower = host.lower()
+        if lower in ("localhost", "metadata", "metadata.google.internal"):
+            return False
+        if lower.endswith((".local", ".localhost", ".internal", ".lan", ".intranet")):
+            return False
+        try:
+            return not _is_private_address(ipaddress.ip_address(host))
+        except ValueError:
+            pass
+        addrs = resolver(host)
+        return bool(addrs) and not any(_is_private_address(a) for a in addrs)
+    except Exception:
+        return False
+
+
+def _resolve_public_ips(
+    url: str,
+    *,
+    resolver: Callable[[str], list[ipaddress._BaseAddress]] | None = None,
+) -> list[ipaddress._BaseAddress]:
+    resolver = resolver or _resolve_hostname_ips
+    parsed = urlparse(url)
+    if parsed.scheme not in ("http", "https") or not parsed.hostname:
+        raise httpx.RequestError(f"Blocked non-public URL: {url}")
+    host = (parsed.hostname or "").strip().lower()
+    if host in ("localhost", "metadata", "metadata.google.internal"):
+        raise httpx.RequestError(f"Blocked non-public hostname: {host}")
+    try:
+        ip = ipaddress.ip_address(host)
+        if _is_private_address(ip):
+            raise httpx.RequestError(f"Blocked non-public IP literal: {host}")
+        return [ip]
+    except httpx.RequestError:
+        raise
+    except ValueError:
+        pass
+    addrs = resolver(host)
+    if not addrs or any(_is_private_address(a) for a in addrs):
+        raise httpx.RequestError(f"Blocked non-public URL: {url}")
+    return addrs
+
+
+class _PinnedBackend(httpcore.NetworkBackend):
+    """Network backend that connects to a pre-resolved IP."""
+
+    def __init__(self, ip: ipaddress._BaseAddress):
+        self._ip = str(ip)
+        self._real = httpcore.SyncBackend()
+
+    def connect_tcp(
+        self,
+        host: str,
+        port: int,
+        timeout: float | None = None,
+        local_address: str | None = None,
+        socket_options=None,
+    ):
+        return self._real.connect_tcp(
+            self._ip, port, timeout, local_address, socket_options
+        )
+
+    def connect_unix_socket(self, path, timeout=None, socket_options=None):
+        return self._real.connect_unix_socket(path, timeout, socket_options)
+
+    def sleep(self, seconds: float) -> None:
+        return self._real.sleep(seconds)
+
+
+_HTTPCORE_TO_HTTPX_EXC = {
+    httpcore.ConnectError: httpx.ConnectError,
+    httpcore.ConnectTimeout: httpx.ConnectTimeout,
+    httpcore.LocalProtocolError: httpx.LocalProtocolError,
+    httpcore.NetworkError: httpx.NetworkError,
+    httpcore.PoolTimeout: httpx.PoolTimeout,
+    httpcore.ProtocolError: httpx.ProtocolError,
+    httpcore.ProxyError: httpx.ProxyError,
+    httpcore.ReadError: httpx.ReadError,
+    httpcore.ReadTimeout: httpx.ReadTimeout,
+    httpcore.RemoteProtocolError: httpx.RemoteProtocolError,
+    httpcore.TimeoutException: httpx.TimeoutException,
+    httpcore.UnsupportedProtocol: httpx.UnsupportedProtocol,
+    httpcore.WriteError: httpx.WriteError,
+    httpcore.WriteTimeout: httpx.WriteTimeout,
+}
+
+
+class _PinnedTransport(httpx.BaseTransport):
+    """Transport that pins every TCP connect to a pre-resolved IP."""
+
+    def __init__(self, ip: ipaddress._BaseAddress, *, http2: bool = False):
+        self._pool = httpcore.ConnectionPool(
+            ssl_context=ssl.create_default_context(),
+            http1=True,
+            http2=http2,
+            network_backend=_PinnedBackend(ip),
+        )
+
+    def __enter__(self):
+        self._pool.__enter__()
+        return self
+
+    def __exit__(self, exc_type=None, exc_value=None, traceback=None) -> None:
+        self._pool.__exit__(exc_type, exc_value, traceback)
+
+    def handle_request(self, request: httpx.Request) -> httpx.Response:
+        httpcore_req = httpcore.Request(
+            method=request.method,
+            url=httpcore.URL(
+                scheme=request.url.raw_scheme,
+                host=request.url.raw_host,
+                port=request.url.port,
+                target=request.url.raw_path,
+            ),
+            headers=request.headers.raw,
+            content=request.stream,
+            extensions=request.extensions,
+        )
+        try:
+            httpcore_resp = self._pool.handle_request(httpcore_req)
+            content = b"".join(cast(Iterable[bytes], httpcore_resp.stream))
+        except Exception as exc:
+            mapped = _HTTPCORE_TO_HTTPX_EXC.get(type(exc))
+            if mapped is not None:
+                raise mapped(str(exc)) from exc
+            raise
+
+        return httpx.Response(
+            status_code=httpcore_resp.status,
+            headers=httpcore_resp.headers,
+            content=content,
+            extensions=httpcore_resp.extensions,
+        )
+
+    def close(self) -> None:
+        self._pool.close()
+
+
+class BodyTooLargeError(Exception):
+    """The server declared a body larger than the hard fetch ceiling."""
+
+    def __init__(self, url: str, declared_bytes: int):
+        self.url = url
+        self.declared_bytes = declared_bytes
+        super().__init__(
+            f"response body is {declared_bytes:,} bytes, over the "
+            f"{WEB_FETCH_HARD_MAX_BYTES:,}-byte hard cap"
+        )
+
+
+class _CappedFetch:
+    """Result of a size-capped streaming GET."""
+
+    __slots__ = (
+        "status_code",
+        "headers",
+        "content",
+        "truncated",
+        "declared_bytes",
+        "encoding",
+        "url",
+    )
+
+    def __init__(
+        self,
+        status_code,
+        headers,
+        content,
+        truncated,
+        declared_bytes,
+        encoding,
+        url,
+    ):
+        self.status_code = status_code
+        self.headers = headers
+        self.content = content
+        self.truncated = truncated
+        self.declared_bytes = declared_bytes
+        self.encoding = encoding
+        self.url = url
+
+    @property
+    def text(self) -> str:
+        return self.content.decode(self.encoding or "utf-8", errors="replace")
+
+    def raise_for_status(self):
+        if self.status_code >= 400:
+            request = httpx.Request("GET", self.url)
+            raise httpx.HTTPStatusError(
+                f"HTTP {self.status_code} for {self.url}",
+                request=request,
+                response=httpx.Response(self.status_code, request=request),
+            )
+
+
+def _get_public_url(
+    url: str,
+    headers: dict,
+    timeout: int,
+    max_redirects: int = 5,
+    max_bytes: int | None = None,
+    *,
+    resolve_public_ips: Callable[[str], list[ipaddress._BaseAddress]] | None = None,
+    transport_factory: Callable[[ipaddress._BaseAddress], httpx.BaseTransport] | None = None,
+) -> _CappedFetch:
+    """Capped streaming GET with SSRF-guarded, DNS-pinned redirects."""
+    resolve_public_ips = resolve_public_ips or _resolve_public_ips
+    transport_factory = transport_factory or _PinnedTransport
+    cap = min(max_bytes or WEB_FETCH_SOFT_MAX_BYTES, WEB_FETCH_HARD_MAX_BYTES)
+    current = url
+    for _ in range(max_redirects + 1):
+        ips = resolve_public_ips(current)
+        req_headers = dict(headers or {})
+        req_headers["Accept-Encoding"] = "identity"
+
+        with httpx.Client(
+            headers=req_headers,
+            timeout=timeout,
+            follow_redirects=False,
+            transport=transport_factory(ips[0]),
+        ) as client:
+            with client.stream("GET", current) as response:
+                if response.status_code in (301, 302, 303, 307, 308):
+                    location = response.headers.get("location")
+                    if not location:
+                        return _CappedFetch(
+                            response.status_code,
+                            response.headers,
+                            b"",
+                            False,
+                            None,
+                            response.encoding,
+                            str(response.url),
+                        )
+                    current = urljoin(str(response.url), location)
+                    continue
+
+                enc = (response.headers.get("content-encoding") or "").strip().lower()
+                if enc and enc != "identity":
+                    raise httpx.RequestError(
+                        f"Refusing compressed response (Content-Encoding: {enc}) after "
+                        "requesting identity: cannot bound decoded body size",
+                        request=httpx.Request("GET", current),
+                    )
+
+                declared = None
+                raw_len = response.headers.get("content-length")
+                if raw_len and raw_len.isdigit():
+                    declared = int(raw_len)
+
+                if declared is not None and declared > WEB_FETCH_HARD_MAX_BYTES:
+                    raise BodyTooLargeError(current, declared)
+
+                chunks = []
+                read = 0
+                truncated = False
+                for chunk in response.iter_bytes():
+                    read += len(chunk)
+                    if read > cap:
+                        keep = cap - (read - len(chunk))
+                        if keep > 0:
+                            chunks.append(chunk[:keep])
+                        truncated = True
+                        break
+                    chunks.append(chunk)
+
+                return _CappedFetch(
+                    response.status_code,
+                    response.headers,
+                    b"".join(chunks),
+                    truncated,
+                    declared,
+                    response.encoding,
+                    str(response.url),
+                )
+
+    raise httpx.RequestError(
+        "Too many redirects", request=httpx.Request("GET", current)
+    )

src/owner_identity.py

@@ -0,0 +1,56 @@
+"""Shared owner identity constants and helpers."""
+
+from __future__ import annotations
+
+import os
+from typing import Optional
+
+
+DEFAULT_LOCAL_OWNER = "__odysseus_local__"
+DEFAULT_LOCAL_OWNER_LABEL = "Local"
+INTERNAL_TOOL_USER = "internal-tool"
+
+REQUEST_SENTINEL_OWNERS = frozenset({INTERNAL_TOOL_USER, "api", "demo", "system"})
+RESERVED_AUTH_USERNAMES = REQUEST_SENTINEL_OWNERS | {DEFAULT_LOCAL_OWNER}
+
+
+def auth_disabled() -> bool:
+    """Return True only when auth is explicitly disabled by configuration."""
+    return os.getenv("AUTH_ENABLED", "true").strip().lower() == "false"
+
+
+def normalize_owner(owner: str | None) -> Optional[str]:
+    """Normalize an owner-like value without inventing a fallback identity."""
+    value = str(owner or "").strip()
+    return value or None
+
+
+def owner_key(owner: str | None) -> Optional[str]:
+    normalized = normalize_owner(owner)
+    return normalized.lower() if normalized else None
+
+
+def is_request_sentinel_owner(owner: str | None) -> bool:
+    return owner_key(owner) in REQUEST_SENTINEL_OWNERS
+
+
+def effective_storage_owner(owner: str | None, *, auth_is_disabled: bool | None = None) -> Optional[str]:
+    """Resolve the owner used for storage writes that need a real bucket.
+
+    ``None`` still means no authenticated owner when auth is enabled. In the
+    explicit no-login mode, it resolves to the reserved local owner instead of
+    conflating local-operator writes with legacy NULL/ownerless rows.
+    """
+    normalized = normalize_owner(owner)
+    if normalized:
+        if is_request_sentinel_owner(normalized):
+            return None
+        return normalized
+    disabled = auth_disabled() if auth_is_disabled is None else auth_is_disabled
+    if disabled:
+        return DEFAULT_LOCAL_OWNER
+    return None
+
+
+def is_default_local_owner(owner: str | None) -> bool:
+    return owner_key(owner) == DEFAULT_LOCAL_OWNER

src/prompt_security.py

@@ -61,7 +61,13 @@ def _sanitize_label(label: str) -> str:
     return label
 
 
-def untrusted_context_message(label: str, content: Any) -> Dict[str, Any]:
+def untrusted_context_message(
+    label: str,
+    content: Any,
+    *,
+    provenance_origin: str | None = None,
+    arm_tool_gate: bool = True,
+) -> Dict[str, Any]:
     """Return an LLM message that keeps retrieved/source text out of system role.
 
     The template is structured so that *only* the hardcoded
@@ -73,6 +79,13 @@ def untrusted_context_message(label: str, content: Any) -> Dict[str, Any]:
     safe_label = _sanitize_label(label)
     text = "" if content is None else str(content)
     text = _escape_guard_markers(text)
+    metadata: Dict[str, Any] = {
+        "trusted": False,
+        "source": label,
+        "tool_gate_untrusted": bool(arm_tool_gate),
+    }
+    if provenance_origin:
+        metadata["provenance_origin"] = provenance_origin
     return {
         "role": "user",
         "content": (
@@ -82,5 +95,5 @@ def untrusted_context_message(label: str, content: Any) -> Dict[str, Any]:
             f"{text}\n"
             f"{GUARD_CLOSE}"
         ),
-        "metadata": {"trusted": False, "source": label},
+        "metadata": metadata,
     }

src/request_models.py

@@ -12,6 +12,7 @@ class ChatRequest(BaseModel):
     use_research: Optional[bool] = Field(default=False, description="Enable deep research")
     time_filter: Optional[str] = Field(default=None, description="Time filter for search")
     preset_id: Optional[str] = Field(default=None, description="Preset identifier")
+    selected_endpoint_id: Optional[str] = Field(default=None, description="Selected model endpoint ID")
     
     @field_validator('message')
     @classmethod

src/settings.py

@@ -14,6 +14,13 @@ from src.constants import SETTINGS_FILE, FEATURES_FILE
 
 logger = logging.getLogger(__name__)
 
+# Keys retained in the raw settings store for compatibility and rollback, but
+# deliberately unavailable through generic settings APIs or agent tools.  They
+# must stay in ``DEFAULT_SETTINGS`` so old files continue to load without data
+# loss; callers that present or mutate settings should use this set as a
+# tombstone boundary.
+RETIRED_SETTING_KEYS = frozenset({"default_model_fallbacks"})
+
 # Tiny TTL cache for settings/features. get_setting() is called on hot paths
 # (every chat, every preprocess); without this it re-parses the JSON each call.
 # Picks up edits within _CACHE_TTL seconds, which is fine for human-edited config.
@@ -197,6 +204,17 @@ DEFAULT_SETTINGS = {
     },
 }
 
+
+def without_retired_settings(settings: dict) -> dict:
+    """Return a shallow copy suitable for generic settings interfaces."""
+    if not isinstance(settings, dict):
+        return {}
+    return {
+        key: value
+        for key, value in settings.items()
+        if key not in RETIRED_SETTING_KEYS
+    }
+
 DEFAULT_FEATURES = {
     "web_search": True,
     "web_fetch": True,
@@ -269,7 +287,7 @@ _PER_USER_KEYS = {
     # Default chat endpoint / model — without per-user resolution every new
     # account inherited whatever the most-recent admin picked, which then
     # got injected into the chat composer on first open.
-    "default_endpoint_id", "default_model", "default_model_fallbacks",
+    "default_endpoint_id", "default_model",
     "utility_endpoint_id", "utility_model", "utility_model_fallbacks",
     "research_endpoint_id", "research_model",
 }

src/task_endpoint.py

@@ -1,7 +1,6 @@
 """Shared resolver for background-task AI endpoints."""
 
 from src.endpoint_resolver import (
-    resolve_chat_fallback_candidates,
     resolve_endpoint,
     resolve_utility_fallback_candidates,
 )
@@ -32,7 +31,6 @@ def resolve_task_candidates(
     2. Utility endpoint/model
     3. Default endpoint/model
     4. Utility fallback chain
-    5. Retired default-fallback compatibility hook (currently empty)
     """
     candidates = []
 
@@ -49,9 +47,6 @@ def resolve_task_candidates(
     _append(*resolve_endpoint("default", owner=owner))
     for url, model, headers in resolve_utility_fallback_candidates(owner=owner):
         _append(url, model, headers)
-    for url, model, headers in resolve_chat_fallback_candidates(owner=owner):
-        _append(url, model, headers)
-
     return candidates
 
 

src/task_scheduler.py

@@ -10,6 +10,7 @@ from datetime import datetime, timedelta, timezone
 from typing import Any, Awaitable, Callable, Dict, Tuple
 
 from core.auth import RESERVED_USERNAMES
+from src.owner_identity import REQUEST_SENTINEL_OWNERS
 from src.task_action_policy import (
     is_admin_only_task_action,
     owner_has_admin_task_privileges,
@@ -1883,6 +1884,7 @@ class TaskScheduler:
             pass
         full_text = ""
         tool_results = []
+        approval_pause = None
 
         # Honor per-task max_steps (defense against runaway agent loops).
         # Falls back to 20 if not set — the historical default.
@@ -1929,9 +1931,44 @@ class TaskScheduler:
                         tool_summary = data.get("stdout") or data.get("output") or data.get("result") or ""
                         if isinstance(tool_summary, str) and tool_summary.strip():
                             tool_results.append(f"[{data.get('tool', '?')}] {tool_summary[:500]}")
+                        approval = data.get("ask_user")
+                        if (
+                            isinstance(approval, dict)
+                            and approval.get("kind") == "tool_approval"
+                        ):
+                            approval_pause = {
+                                "tool": data.get("tool") or "tool",
+                                "approval_id": approval.get("approval_id"),
+                            }
+                            # Scheduled tasks have no interactive surface that
+                            # can safely resume a one-use grant. Retire the
+                            # record immediately instead of leaving it pending
+                            # and report an explicit manual-action boundary.
+                            try:
+                                from src.tool_approvals import tool_approval_store
+                                tool_approval_store.consume(
+                                    approval_pause["approval_id"],
+                                    decision="deny",
+                                    owner=task.owner,
+                                    session_id=session_id,
+                                )
+                            except Exception:
+                                logger.debug(
+                                    "Could not retire scheduled-task approval",
+                                    exc_info=True,
+                                )
+                            break
                 except (json.JSONDecodeError, KeyError):
                     pass
 
+        if approval_pause is not None:
+            return (
+                "Scheduled task paused safely: "
+                f"{approval_pause['tool']} requested an exact action after "
+                "untrusted context. That action was not executed. Run this task "
+                "interactively to inspect and approve the action."
+            )
+
         # Grace summarization — if the model exhausted rounds on tool calls
         # without producing a final text response, do one last LLM call
         # asking it to summarize what it did. Guarantees output.
@@ -2484,7 +2521,7 @@ class TaskScheduler:
         # check-ins seeded, which then double-fire alongside the human user's
         # check-ins. This was the root cause of the duplicate 'Morning check-in'
         # rows we had to manually clean up.
-        if not owner or owner in RESERVED_USERNAMES:
+        if not owner or owner in REQUEST_SENTINEL_OWNERS:
             logger.info(f"ensure_assistant_defaults: skip synthetic owner {owner!r}")
             return
         from core.database import SessionLocal, CrewMember, ScheduledTask

src/teacher_escalation.py

@@ -439,56 +439,11 @@ async def escalate_and_learn(
     failure_reason: str,
     owner: Optional[str] = None,
 ) -> Optional[str]:
-    """Call the teacher, evaluate ITS attempt, save a skill on success.
-
-    Returns the saved skill name (or None if the teacher couldn't
-    write one). Logs but doesn't raise — escalation is best-effort.
-    """
-    from src.settings import get_setting
-    teacher_spec = (get_setting("teacher_model", "") or "").strip()
-    if not teacher_spec:
-        return None
-
-    prompt = _TEACHER_ESCALATION_PROMPT.format(
-        user_request=user_request or "(no user request captured)",
-        failure_reason=failure_reason or "(failure reason not captured)",
-        untrusted_trace_guard=_UNTRUSTED_TRACE_GUARD,
-        trace=_format_trace(tool_results, agent_reply),
+    """Retire legacy background learning when no approval UI is available."""
+    logger.info(
+        "background teacher learning skipped: generated skills require an "
+        "interactive exact approval"
     )
-    response = await _call_teacher(teacher_spec, prompt, owner=owner)
-    if not response:
-        return None
-
-    skill = _extract_skill_json(response)
-    if not skill:
-        # Teacher chose not to write a skill — see prompt contract.
-        logger.info("teacher declined to write a skill for this failure")
-        return None
-
-    # Same regex eval applied to the teacher's response — if the
-    # teacher itself sounded uncertain ("I don't have a tool"), drop
-    # the skill rather than persist a sketchy one.
-    status, reason = evaluate_turn_regex([], response)
-    if status == "failure":
-        logger.info(f"teacher response failed eval, skipping skill save: {reason}")
-        return None
-
-    # Tag the skill with the escalation source for auditability.
-    skill.setdefault("source", "teacher-escalation")
-    skill.setdefault("teacher_model", teacher_spec)
-    # Force action=add regardless of what the teacher wrote.
-    skill["action"] = "add"
-
-    import json
-    from src.tool_implementations import do_manage_skills
-    try:
-        result = await do_manage_skills(json.dumps(skill), owner=owner)
-        if isinstance(result, dict) and not result.get("error"):
-            logger.info(f"teacher wrote skill: {skill.get('name')}")
-            return skill.get("name")
-        logger.warning(f"skill save failed: {result}")
-    except Exception as e:
-        logger.warning(f"skill save raised: {e}")
     return None
 
 
@@ -563,6 +518,12 @@ async def run_teacher_inline(
     student_tool_events: List[Dict[str, Any]],
     student_reply: str,
     owner: Optional[str] = None,
+    session_id: Optional[str] = None,
+    workspace: Optional[str] = None,
+    disabled_tools: Optional[set[str]] = None,
+    tool_policy: Any = None,
+    active_document: Any = None,
+    active_email: Optional[Dict[str, str]] = None,
 ):
     """Async generator. Yields SSE event strings.
 
@@ -661,6 +622,7 @@ async def run_teacher_inline(
     from src.agent_loop import stream_agent_loop
     captured_tool_events: List[Dict[str, Any]] = []
     captured_text_parts: List[str] = []
+    captured_metrics: Dict[str, Any] = {}
 
     async for evt_str in stream_agent_loop(
         endpoint_url=teacher_url,
@@ -668,6 +630,12 @@ async def run_teacher_inline(
         messages=teacher_messages,
         headers=teacher_headers,
         owner=owner,
+        session_id=session_id,
+        workspace=workspace,
+        disabled_tools=disabled_tools,
+        tool_policy=tool_policy,
+        active_document=active_document,
+        active_email=active_email,
         _is_teacher_run=True,
     ):
         # Swallow teacher's own [DONE] — outer loop emits the real one
@@ -682,13 +650,21 @@ async def run_teacher_inline(
             if isinstance(payload, dict):
                 payload["teacher"] = True
                 typ = payload.get("type")
+                if typ == "metrics" and isinstance(payload.get("data"), dict):
+                    # The outer chat route persists only the last metrics
+                    # payload. Keep a copy so any approval produced after the
+                    # recursive teacher run's metrics remains reloadable.
+                    captured_metrics = dict(payload["data"])
                 if typ == "tool_output":
-                    captured_tool_events.append({
+                    captured_tool_event = {
                         "tool": payload.get("tool"),
                         "command": payload.get("command"),
                         "output": payload.get("output"),
                         "exit_code": payload.get("exit_code"),
-                    })
+                    }
+                    if isinstance(payload.get("ask_user"), dict):
+                        captured_tool_event["ask_user"] = payload["ask_user"]
+                    captured_tool_events.append(captured_tool_event)
                 if "delta" in payload and isinstance(payload["delta"], str):
                     if payload.get("thinking"):
                         continue
@@ -697,6 +673,12 @@ async def run_teacher_inline(
                 continue
         yield evt_str
 
+    # A takeover that paused for a question or exact action has not completed
+    # yet. Its server-owned approval card is already in the live/persisted tool
+    # events; do not evaluate the partial trace or distill it into a skill.
+    if any(event.get("ask_user") for event in captured_tool_events):
+        return
+
     teacher_text = "".join(captured_text_parts).strip()
     t_status, t_reason = evaluate_turn_regex(captured_tool_events, teacher_text)
     if t_status == "failure":
@@ -740,31 +722,85 @@ async def run_teacher_inline(
     skill.setdefault("source", "teacher-escalation")
     skill.setdefault("teacher_model", teacher_spec)
 
-    import json as _json
-    from src.tool_implementations import do_manage_skills
-    try:
-        result = await do_manage_skills(_json.dumps(skill), owner=owner)
-        if isinstance(result, dict) and not result.get("error"):
-            logger.info(f"teacher succeeded; saved skill: {skill.get('name')}")
-            yield (
-                'data: ' + json.dumps({
-                    "type": "skill_saved",
-                    "name": skill.get("name"),
-                    "category": skill.get("category", "general"),
-                }) + '\n\n'
-            )
-        else:
-            yield (
-                'data: ' + json.dumps({
-                    "type": "skill_save_failed",
-                    "reason": str(result),
-                }) + '\n\n'
-            )
-    except Exception as e:
-        logger.warning(f"skill save raised: {e}")
+    if not session_id:
         yield (
             'data: ' + json.dumps({
                 "type": "skill_save_failed",
-                "reason": str(e),
+                "reason": (
+                    "Teacher-generated skills require an interactive exact "
+                    "approval before they can be saved."
+                ),
             }) + '\n\n'
         )
+        return
+
+    import json as _json
+    import uuid as _uuid
+    from src.tool_approvals import tool_approval_store
+    from src.tool_capabilities import capabilities_for_action
+
+    skill_content = _json.dumps(skill, ensure_ascii=False)
+    pending = tool_approval_store.create(
+        owner=owner,
+        session_id=session_id,
+        origin_run_id=f"teacher-skill-{_uuid.uuid4().hex}",
+        tool_name="manage_skills",
+        content=skill_content,
+        workspace=workspace,
+        external_untrusted_context_seen=True,
+        capabilities=capabilities_for_action("manage_skills", skill_content),
+    )
+    approval = pending.public_payload(
+        reason=(
+            "The teacher generated this reusable skill. Review and approve "
+            "the complete skill definition before it is saved."
+        ),
+    )
+    persisted_metrics = dict(captured_metrics)
+    persisted_tool_events = list(persisted_metrics.get("tool_events") or [])
+    persisted_round_texts = list(persisted_metrics.get("round_texts") or [])
+    prior_rounds = [
+        event.get("round")
+        for event in persisted_tool_events
+        if isinstance(event, dict) and isinstance(event.get("round"), int)
+    ]
+    approval_round = max([len(persisted_round_texts), *prior_rounds, 0]) + 1
+    approval_tool_event = {
+        "round": approval_round,
+        "model": teacher_model,
+        "tool": "manage_skills",
+        "command": str(skill.get("name") or "teacher-generated skill"),
+        "output": "Waiting for an exact user approval.",
+        "exit_code": None,
+        "ask_user": approval,
+    }
+    persisted_tool_events.append(approval_tool_event)
+    persisted_metrics["tool_events"] = persisted_tool_events
+    persisted_metrics.setdefault("model", teacher_model)
+    yield (
+        "data: "
+        + json.dumps({"delta": "Review the teacher-generated skill before saving it."})
+        + "\n\n"
+    )
+    yield (
+        "data: "
+        + json.dumps({
+            "type": "tool_output",
+            **approval_tool_event,
+            "teacher": True,
+        })
+        + "\n\n"
+    )
+    yield (
+        "data: "
+        + json.dumps({"type": "ask_user", "data": approval, "teacher": True})
+        + "\n\n"
+    )
+    # This must be the final metrics event: chat_routes saves only last_metrics
+    # when the outer stream reaches [DONE]. Without it, the live approval card
+    # disappears after a reload even though the server grant remains pending.
+    yield (
+        "data: "
+        + json.dumps({"type": "metrics", "data": persisted_metrics, "teacher": True})
+        + "\n\n"
+    )

src/tool_approvals.py

@@ -0,0 +1,392 @@
+"""Opaque, exact, one-use approvals for tainted model-requested actions.
+
+The model may propose an action after untrusted context, but only the server
+stores and later executes the exact approved tool input.  Browser-visible
+fields are display copies, never authority.
+"""
+
+from __future__ import annotations
+
+import hashlib
+import json
+import os
+import secrets
+import threading
+import time
+from dataclasses import dataclass, field
+from typing import Any
+
+from src.tool_capabilities import ToolCapabilities, capabilities_for_action
+
+
+DEFAULT_APPROVAL_TTL_SECONDS = 10 * 60
+DEFAULT_MAX_PENDING_APPROVALS = 2048
+
+
+def _normalized_owner(owner: Any) -> str:
+    return str(owner or "").strip().casefold()
+
+
+def _normalized_workspace(workspace: Any) -> str:
+    if not isinstance(workspace, str) or not workspace.strip():
+        return ""
+    return os.path.realpath(os.path.expanduser(workspace))
+
+
+def _canonical_digest(payload: dict[str, Any]) -> str:
+    encoded = json.dumps(
+        payload,
+        sort_keys=True,
+        separators=(",", ":"),
+        ensure_ascii=False,
+    ).encode("utf-8")
+    return hashlib.sha256(encoded).hexdigest()
+
+
+def document_content_digest(content: Any) -> str:
+    """Return the stable server-side fingerprint used to seal a document."""
+    return hashlib.sha256(str(content or "").encode("utf-8")).hexdigest()
+
+
+def _binding_payload(
+    *,
+    owner: Any,
+    session_id: Any,
+    origin_run_id: Any,
+    tool_name: Any,
+    content: Any,
+    workspace: Any,
+    document_id: Any,
+    document_version: Any,
+    document_digest: Any,
+    external_untrusted_context_seen: bool,
+    effects: tuple[str, ...],
+    result_integrity: str,
+) -> dict[str, Any]:
+    return {
+        "owner": _normalized_owner(owner),
+        "session_id": str(session_id or ""),
+        "origin_run_id": str(origin_run_id or ""),
+        "tool_name": str(tool_name or ""),
+        "content": str(content or ""),
+        "workspace": _normalized_workspace(workspace),
+        "document_id": str(document_id or ""),
+        "document_version": (
+            int(document_version) if document_version is not None else None
+        ),
+        "document_digest": str(document_digest or "").strip().lower(),
+        "external_untrusted_context_seen": bool(external_untrusted_context_seen),
+        "effects": list(effects),
+        "result_integrity": str(result_integrity),
+    }
+
+
+@dataclass(frozen=True)
+class PendingToolApproval:
+    approval_id: str
+    owner: str
+    session_id: str
+    origin_run_id: str
+    tool_name: str
+    content: str
+    workspace: str
+    document_id: str
+    document_version: int | None
+    document_digest: str
+    external_untrusted_context_seen: bool
+    effects: tuple[str, ...]
+    result_integrity: str
+    digest: str
+    created_at: float
+    expires_at: float
+
+    def public_payload(self, *, reason: str | None = None) -> dict[str, Any]:
+        return {
+            "kind": "tool_approval",
+            "approval_id": self.approval_id,
+            "question": "Allow this exact action once?",
+            "description": reason or (
+                "Untrusted context influenced this run, so this action needs "
+                "your explicit approval."
+            ),
+            "options": [
+                {
+                    "label": "Allow once",
+                    "value": "approve",
+                    "description": "Execute only the sealed action shown here.",
+                },
+                {
+                    "label": "Deny",
+                    "value": "deny",
+                    "description": "Do not execute it.",
+                },
+            ],
+            "action": {
+                "tool": self.tool_name,
+                # Show the complete sealed input so approval never hides
+                # trailing lines.  This is not read back as authority.
+                "content": self.content,
+                "digest": self.digest[:16],
+                "effects": list(self.effects),
+                "workspace": self.workspace or None,
+                "document_id": self.document_id or None,
+                "document_version": self.document_version,
+            },
+        }
+
+
+@dataclass
+class ExactToolApproval:
+    """A consumed grant that the dispatcher can claim exactly once."""
+
+    pending: PendingToolApproval
+    _claimed: bool = field(default=False, init=False, repr=False)
+    _lock: threading.Lock = field(default_factory=threading.Lock, init=False, repr=False)
+
+    def _matches_unlocked(
+        self,
+        *,
+        owner: Any,
+        session_id: Any,
+        tool_name: Any,
+        content: Any,
+        workspace: Any,
+    ) -> bool:
+        if self._claimed:
+            return False
+        capabilities = capabilities_for_action(tool_name, content)
+        effects = tuple(sorted(effect.value for effect in capabilities.effects))
+        result_integrity = capabilities.result_integrity.value
+        if (
+            effects != self.pending.effects
+            or result_integrity != self.pending.result_integrity
+        ):
+            return False
+        expected = _binding_payload(
+            owner=owner,
+            session_id=session_id,
+            origin_run_id=self.pending.origin_run_id,
+            tool_name=tool_name,
+            content=content,
+            workspace=workspace,
+            document_id=self.pending.document_id,
+            document_version=self.pending.document_version,
+            document_digest=self.pending.document_digest,
+            external_untrusted_context_seen=(
+                self.pending.external_untrusted_context_seen
+            ),
+            effects=effects,
+            result_integrity=result_integrity,
+        )
+        return _canonical_digest(expected) == self.pending.digest
+
+    def matches(
+        self,
+        *,
+        owner: Any,
+        session_id: Any,
+        tool_name: Any,
+        content: Any,
+        workspace: Any,
+    ) -> bool:
+        with self._lock:
+            return self._matches_unlocked(
+                owner=owner,
+                session_id=session_id,
+                tool_name=tool_name,
+                content=content,
+                workspace=workspace,
+            )
+
+    def claim(
+        self,
+        *,
+        owner: Any,
+        session_id: Any,
+        tool_name: Any,
+        content: Any,
+        workspace: Any,
+    ) -> bool:
+        with self._lock:
+            if not self._matches_unlocked(
+                owner=owner,
+                session_id=session_id,
+                tool_name=tool_name,
+                content=content,
+                workspace=workspace,
+            ):
+                return False
+            self._claimed = True
+            return True
+
+
+class ToolApprovalStore:
+    """Thread-safe pending approval registry with destructive consumption."""
+
+    def __init__(
+        self,
+        *,
+        ttl_seconds: int = DEFAULT_APPROVAL_TTL_SECONDS,
+        max_pending: int = DEFAULT_MAX_PENDING_APPROVALS,
+    ):
+        self._ttl_seconds = max(1, int(ttl_seconds))
+        self._max_pending = max(1, int(max_pending))
+        self._pending: dict[str, PendingToolApproval] = {}
+        self._lock = threading.Lock()
+
+    def _purge_expired_locked(self, now: float) -> None:
+        expired = [
+            approval_id
+            for approval_id, pending in self._pending.items()
+            if pending.expires_at <= now
+        ]
+        for approval_id in expired:
+            self._pending.pop(approval_id, None)
+
+    def create(
+        self,
+        *,
+        owner: Any,
+        session_id: Any,
+        origin_run_id: Any,
+        tool_name: Any,
+        content: Any,
+        workspace: Any,
+        document_id: Any = None,
+        document_version: Any = None,
+        document_digest: Any = None,
+        external_untrusted_context_seen: bool,
+        capabilities: ToolCapabilities,
+    ) -> PendingToolApproval:
+        now = time.time()
+        effects = tuple(sorted(effect.value for effect in capabilities.effects))
+        result_integrity = capabilities.result_integrity.value
+        payload = _binding_payload(
+            owner=owner,
+            session_id=session_id,
+            origin_run_id=origin_run_id,
+            tool_name=tool_name,
+            content=content,
+            workspace=workspace,
+            document_id=document_id,
+            document_version=document_version,
+            document_digest=document_digest,
+            external_untrusted_context_seen=external_untrusted_context_seen,
+            effects=effects,
+            result_integrity=result_integrity,
+        )
+        pending = PendingToolApproval(
+            approval_id=secrets.token_urlsafe(32),
+            owner=payload["owner"],
+            session_id=payload["session_id"],
+            origin_run_id=payload["origin_run_id"],
+            tool_name=payload["tool_name"],
+            content=payload["content"],
+            workspace=payload["workspace"],
+            document_id=payload["document_id"],
+            document_version=payload["document_version"],
+            document_digest=payload["document_digest"],
+            external_untrusted_context_seen=payload[
+                "external_untrusted_context_seen"
+            ],
+            effects=effects,
+            result_integrity=result_integrity,
+            digest=_canonical_digest(payload),
+            created_at=now,
+            expires_at=now + self._ttl_seconds,
+        )
+        with self._lock:
+            self._purge_expired_locked(now)
+            # The chat UI exposes one pending card per session, so supersede an
+            # older action there. Headless/manual-test callers use an empty
+            # session id; keep independent origin runs separate so two skill
+            # tests owned by the same user cannot invalidate each other.
+            superseded = [
+                approval_id
+                for approval_id, existing in self._pending.items()
+                if (
+                    existing.owner == pending.owner
+                    and existing.session_id == pending.session_id
+                    and (
+                        bool(pending.session_id)
+                        or existing.origin_run_id == pending.origin_run_id
+                    )
+                )
+            ]
+            for approval_id in superseded:
+                self._pending.pop(approval_id, None)
+            while len(self._pending) >= self._max_pending:
+                oldest_id = min(
+                    self._pending,
+                    key=lambda approval_id: self._pending[approval_id].created_at,
+                )
+                self._pending.pop(oldest_id, None)
+            self._pending[pending.approval_id] = pending
+        return pending
+
+    def consume(
+        self,
+        approval_id: Any,
+        *,
+        decision: Any,
+        owner: Any,
+        session_id: Any,
+    ) -> ExactToolApproval | None:
+        now = time.time()
+        with self._lock:
+            self._purge_expired_locked(now)
+            approval_key = str(approval_id or "")
+            pending = self._pending.get(approval_key)
+            if pending is None:
+                return None
+            if (
+                pending.owner != _normalized_owner(owner)
+                or pending.session_id != str(session_id or "")
+            ):
+                # Authentication is checked before destructive consumption so
+                # a leaked/guessed opaque id cannot be used to invalidate
+                # another owner's pending action.
+                return None
+            self._pending.pop(approval_key, None)
+        if str(decision or "").strip().lower() != "approve":
+            return None
+        return ExactToolApproval(pending)
+
+    def peek(self, approval_id: Any) -> PendingToolApproval | None:
+        now = time.time()
+        with self._lock:
+            self._purge_expired_locked(now)
+            return self._pending.get(str(approval_id or ""))
+
+    def retire_for_session(self, *, owner: Any, session_id: Any) -> bool:
+        """Discard pending actions superseded by an ordinary user turn.
+
+        Returns whether any retired action carried external provenance, so the
+        caller can preserve that security state without treating the new user
+        message as an approval continuation.
+        """
+        now = time.time()
+        normalized_owner = _normalized_owner(owner)
+        normalized_session = str(session_id or "")
+        if not normalized_session:
+            return False
+        with self._lock:
+            self._purge_expired_locked(now)
+            retired_ids = [
+                approval_id
+                for approval_id, pending in self._pending.items()
+                if (
+                    pending.owner == normalized_owner
+                    and pending.session_id == normalized_session
+                )
+            ]
+            carried_taint = any(
+                self._pending[approval_id].external_untrusted_context_seen
+                for approval_id in retired_ids
+            )
+            for approval_id in retired_ids:
+                self._pending.pop(approval_id, None)
+        return carried_taint
+
+
+tool_approval_store = ToolApprovalStore()

src/tool_capabilities.py

@@ -0,0 +1,668 @@
+"""Deterministic capability metadata for agent tools.
+
+Model output requests an action; it never supplies the authority for that
+action.  This module classifies the effects of each built-in tool and applies
+run-local integrity gates before dispatch.
+"""
+
+from __future__ import annotations
+
+import json
+import uuid
+from dataclasses import dataclass, field
+from enum import Enum
+from types import MappingProxyType
+from typing import Any, Iterable, Mapping
+
+from src.tool_security import BUILTIN_EMAIL_TOOLS
+
+
+class ToolEffect(str, Enum):
+    READ_PUBLIC = "read_public"
+    READ_WORKSPACE = "read_workspace"
+    READ_PRIVATE = "read_private"
+    WRITE_WORKSPACE = "write_workspace"
+    WRITE_PRIVATE = "write_private"
+    EXECUTE_CODE = "execute_code"
+    BROKERED_NETWORK_READ = "brokered_network_read"
+    NETWORK_EGRESS = "network_egress"
+    EXTERNAL_SIDE_EFFECT = "external_side_effect"
+    UI_SIDE_EFFECT = "ui_side_effect"
+    ADMIN_CHANGE = "admin_change"
+    DESTRUCTIVE = "destructive"
+    USER_INTERACTION = "user_interaction"
+
+
+class ResultIntegrity(str, Enum):
+    SYSTEM = "system"
+    WORKSPACE_UNTRUSTED = "workspace_untrusted"
+    EXTERNAL_UNTRUSTED = "external_untrusted"
+
+
+@dataclass(frozen=True)
+class ToolCapabilities:
+    effects: frozenset[ToolEffect]
+    result_integrity: ResultIntegrity = ResultIntegrity.SYSTEM
+    known: bool = True
+
+
+def _capabilities(
+    *effects: ToolEffect,
+    result_integrity: ResultIntegrity = ResultIntegrity.SYSTEM,
+) -> ToolCapabilities:
+    return ToolCapabilities(frozenset(effects), result_integrity)
+
+
+_REGISTRY: dict[str, ToolCapabilities] = {}
+
+
+def _register(
+    names: Iterable[str],
+    *effects: ToolEffect,
+    result_integrity: ResultIntegrity = ResultIntegrity.SYSTEM,
+) -> None:
+    capabilities = _capabilities(*effects, result_integrity=result_integrity)
+    for name in names:
+        if name in _REGISTRY:
+            raise RuntimeError(f"Duplicate tool capability classification: {name}")
+        _REGISTRY[name] = capabilities
+
+
+_register(
+    {"ask_user", "update_plan"},
+    ToolEffect.USER_INTERACTION,
+)
+_register(
+    {
+        "list_cached_models",
+        "list_cookbook_servers",
+        "list_downloads",
+        "list_models",
+        "list_serve_presets",
+        "list_served_models",
+    },
+    ToolEffect.READ_PRIVATE,
+    # These readers return provider-controlled model identifiers or durable
+    # user/admin-authored Cookbook and process state.  Local brokering does not
+    # make the returned text server-authored.
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"search_hf_models"},
+    ToolEffect.BROKERED_NETWORK_READ,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"get_workspace", "glob", "grep", "ls", "read_file"},
+    ToolEffect.READ_WORKSPACE,
+    result_integrity=ResultIntegrity.WORKSPACE_UNTRUSTED,
+)
+_register(
+    {"web_search"},
+    ToolEffect.BROKERED_NETWORK_READ,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"web_fetch"},
+    ToolEffect.BROKERED_NETWORK_READ,
+    ToolEffect.NETWORK_EGRESS,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {
+        "list_email_accounts",
+        "list_emails",
+        "read_email",
+        "resolve_contact",
+        "scan_email_unsubscribes",
+        "search_chats",
+        "search_emails",
+        "list_sessions",
+        "tail_serve_output",
+        "vault_get",
+        "vault_search",
+    },
+    ToolEffect.READ_PRIVATE,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"bash", "manage_bg_jobs", "python"},
+    ToolEffect.EXECUTE_CODE,
+    result_integrity=ResultIntegrity.WORKSPACE_UNTRUSTED,
+)
+_register(
+    {"apply_patch", "edit_file", "write_file"},
+    ToolEffect.WRITE_WORKSPACE,
+    # Successful writes include unified diffs that can echo arbitrary existing
+    # workspace content back into the next model round.
+    result_integrity=ResultIntegrity.WORKSPACE_UNTRUSTED,
+)
+_register(
+    {
+        "create_document",
+        "manage_calendar",
+        "manage_contact",
+        "manage_documents",
+        "manage_memory",
+        "manage_notes",
+        "manage_research",
+        "manage_session",
+        "manage_skills",
+        "manage_tasks",
+        "suggest_document",
+        "todowrite",
+    },
+    ToolEffect.WRITE_PRIVATE,
+)
+_register(
+    {
+        "ai_draft_email_reply",
+        "create_session",
+        "draft_email",
+        "draft_email_reply",
+    },
+    ToolEffect.WRITE_PRIVATE,
+    # These tools resolve user-configured endpoints/accounts or read stored
+    # email content before returning model-visible status text.
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"edit_document", "update_document"},
+    ToolEffect.WRITE_PRIVATE,
+    # These tools can echo stored document content that was not present in
+    # their arguments.  edit_document returns the complete edited document;
+    # update_document also preserves stored email headers/thread history.
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"pipeline"},
+    ToolEffect.NETWORK_EGRESS,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"send_to_session"},
+    ToolEffect.NETWORK_EGRESS,
+    ToolEffect.WRITE_PRIVATE,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"chat_with_model", "ask_teacher"},
+    ToolEffect.NETWORK_EGRESS,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"download_attachment"},
+    ToolEffect.READ_PRIVATE,
+    ToolEffect.WRITE_WORKSPACE,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"edit_image", "generate_image", "trigger_research"},
+    ToolEffect.NETWORK_EGRESS,
+    ToolEffect.WRITE_PRIVATE,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {
+        "archive_email",
+        "bulk_email",
+        "mark_email_read",
+        "reply_to_email",
+        "send_email",
+        "unsubscribe_email",
+    },
+    ToolEffect.EXTERNAL_SIDE_EFFECT,
+    # Email action results can include stored headers/account labels or remote
+    # SMTP/IMAP responses, even when the action itself succeeded.
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"delete_email"},
+    ToolEffect.EXTERNAL_SIDE_EFFECT,
+    ToolEffect.DESTRUCTIVE,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {"ui_control"},
+    ToolEffect.UI_SIDE_EFFECT,
+    # Model switches and custom-theme validation read mutable user settings.
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {
+        "adopt_served_model",
+        "cancel_download",
+        "download_model",
+        "serve_model",
+        "serve_preset",
+        "stop_served_model",
+        "vault_unlock",
+    },
+    ToolEffect.ADMIN_CHANGE,
+    # Cookbook/process operations can return stored presets, provider data,
+    # remote shell output, and command errors.
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_register(
+    {
+        "api_call",
+        "app_api",
+        "manage_endpoints",
+        "manage_mcp",
+        "manage_settings",
+        "manage_tokens",
+        "manage_webhooks",
+    },
+    ToolEffect.ADMIN_CHANGE,
+    # api_call/app_api return remote or stored application data, and the
+    # admin managers can echo user-controlled configuration.  Conservatively
+    # retain the action effect while treating every successful result as data.
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+
+
+TOOL_CAPABILITIES: Mapping[str, ToolCapabilities] = MappingProxyType(dict(_REGISTRY))
+KNOWN_CAPABILITY_TOOLS = frozenset(TOOL_CAPABILITIES)
+
+_UNKNOWN_CAPABILITIES = _capabilities(
+    ToolEffect.READ_PRIVATE,
+    ToolEffect.WRITE_WORKSPACE,
+    ToolEffect.WRITE_PRIVATE,
+    ToolEffect.EXECUTE_CODE,
+    ToolEffect.NETWORK_EGRESS,
+    ToolEffect.EXTERNAL_SIDE_EFFECT,
+    ToolEffect.ADMIN_CHANGE,
+    ToolEffect.DESTRUCTIVE,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_UNKNOWN_CAPABILITIES = ToolCapabilities(
+    _UNKNOWN_CAPABILITIES.effects,
+    _UNKNOWN_CAPABILITIES.result_integrity,
+    known=False,
+)
+_BROWSER_MCP_READ_CAPABILITIES = _capabilities(
+    ToolEffect.BROKERED_NETWORK_READ,
+    result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+)
+_BROWSER_MCP_READ_TOOLS = frozenset(
+    {
+        "mcp__builtin_browser__browser_console_messages",
+        "mcp__builtin_browser__browser_network_requests",
+        "mcp__builtin_browser__browser_snapshot",
+        "mcp__builtin_browser__browser_take_screenshot",
+    }
+)
+
+
+def capabilities_for_tool(tool_name: Any) -> ToolCapabilities:
+    """Return deterministic capabilities; malformed and unknown tools fail high."""
+    if not isinstance(tool_name, str) or not tool_name:
+        return _UNKNOWN_CAPABILITIES
+    capabilities = TOOL_CAPABILITIES.get(tool_name)
+    if capabilities is not None:
+        return capabilities
+    if tool_name.startswith("mcp__email__"):
+        bare_name = tool_name[len("mcp__email__"):]
+        capabilities = TOOL_CAPABILITIES.get(bare_name)
+        if bare_name in BUILTIN_EMAIL_TOOLS and capabilities is not None:
+            return capabilities
+    if tool_name in _BROWSER_MCP_READ_TOOLS:
+        return _BROWSER_MCP_READ_CAPABILITIES
+    return _UNKNOWN_CAPABILITIES
+
+
+_PRIVATE_ACTION_READS: Mapping[str, frozenset[str]] = MappingProxyType(
+    {
+        "manage_calendar": frozenset({"list_calendars", "list_events"}),
+        "manage_contact": frozenset({"list"}),
+        "manage_documents": frozenset({"list", "read", "view", "open", "get"}),
+        "manage_memory": frozenset({"list", "search"}),
+        "manage_notes": frozenset({"list", "search", "find", "view"}),
+        "manage_research": frozenset({"list", "read", "open", "view", "get"}),
+        "manage_session": frozenset({"list", "switch", "open", "select", "view"}),
+        "manage_skills": frozenset({"list", "index", "view", "view_ref", "search"}),
+        "manage_tasks": frozenset({"list"}),
+    }
+)
+
+_PRIVATE_ACTION_WRITES: Mapping[str, frozenset[str]] = MappingProxyType(
+    {
+        "manage_calendar": frozenset(
+            {"create_event", "update_event", "delete_event"}
+        ),
+        "manage_contact": frozenset({"add", "update", "edit", "delete"}),
+        "manage_documents": frozenset({"delete", "tidy"}),
+        "manage_memory": frozenset({"add", "edit", "delete"}),
+        "manage_notes": frozenset({"add", "update", "delete", "toggle_item"}),
+        "manage_research": frozenset({"delete"}),
+        "manage_session": frozenset(
+            {
+                "rename",
+                "archive",
+                "unarchive",
+                "delete",
+                "important",
+                "unimportant",
+                "truncate",
+                "fork",
+            }
+        ),
+        "manage_skills": frozenset({"add", "edit", "patch", "publish", "delete"}),
+        "manage_tasks": frozenset({"create", "edit", "delete", "pause", "resume", "run"}),
+    }
+)
+
+_ACTION_DESTRUCTIVE: Mapping[str, frozenset[str]] = MappingProxyType(
+    {
+        "manage_calendar": frozenset({"delete_event"}),
+        "manage_contact": frozenset({"delete"}),
+        "manage_documents": frozenset({"delete", "tidy"}),
+        "manage_endpoints": frozenset({"delete"}),
+        "manage_bg_jobs": frozenset({"kill", "stop", "cancel", "terminate"}),
+        "manage_memory": frozenset({"delete"}),
+        "manage_mcp": frozenset({"delete"}),
+        "manage_notes": frozenset({"delete"}),
+        "manage_research": frozenset({"delete"}),
+        "manage_session": frozenset({"delete", "truncate"}),
+        "manage_settings": frozenset({"delete", "reset"}),
+        "manage_skills": frozenset({"delete"}),
+        "manage_tasks": frozenset({"delete"}),
+        "manage_tokens": frozenset({"delete"}),
+        "manage_webhooks": frozenset({"delete"}),
+    }
+)
+
+_ACTION_DEFAULTS: Mapping[str, str] = MappingProxyType(
+    {
+        "manage_calendar": "list_events",
+        "manage_documents": "list",
+        "manage_research": "list",
+        "manage_tasks": "list",
+    }
+)
+
+_ACTION_ALIASES: Mapping[str, Mapping[str, str]] = MappingProxyType(
+    {
+        "manage_calendar": MappingProxyType(
+            {
+                "create": "create_event",
+                "update": "update_event",
+                "delete": "delete_event",
+                "list": "list_events",
+            }
+        ),
+        "manage_notes": MappingProxyType(
+            {
+                "create": "add",
+                "new": "add",
+                "save": "add",
+                "remind": "add",
+                "reminder": "add",
+                "remove": "delete",
+                "remove_item": "toggle_item",
+            }
+        ),
+    }
+)
+
+_LINE_ACTION_TOOLS = frozenset({"manage_memory", "manage_session"})
+
+
+def _action_from_content(tool_name: str, content: Any) -> str | None:
+    """Extract the action discriminator using the same accepted input shapes."""
+    if isinstance(content, Mapping):
+        payload: Any = dict(content)
+    elif isinstance(content, str):
+        raw = content.strip()
+        if tool_name in _LINE_ACTION_TOOLS and raw and not raw.startswith("{"):
+            return raw.splitlines()[0].strip().replace("-", "_").casefold() or None
+        try:
+            payload = json.loads(raw) if raw else {}
+        except (TypeError, ValueError):
+            return None
+    else:
+        payload = {}
+
+    if not isinstance(payload, dict):
+        return None
+    if (
+        len(payload) == 1
+        and isinstance(payload.get("body"), dict)
+        and "action" in payload["body"]
+    ):
+        payload = payload["body"]
+
+    action = payload.get("action")
+    if (
+        not action
+        and tool_name == "manage_calendar"
+        and isinstance(payload.get("events"), list)
+    ):
+        action = "create_event"
+    if not action and tool_name == "manage_tasks" and any(
+        payload.get(key) is not None
+        for key in ("task", "description", "schedule", "time", "day_of_week")
+    ):
+        action = "create"
+    if not isinstance(action, str) or not action.strip():
+        action = _ACTION_DEFAULTS.get(tool_name)
+    if not action:
+        return None
+    normalized = action.strip().replace("-", "_").casefold()
+    return _ACTION_ALIASES.get(tool_name, {}).get(normalized, normalized)
+
+
+def capabilities_for_action(tool_name: Any, content: Any) -> ToolCapabilities:
+    """Classify a sealed multiplexed action; ambiguous actions fail high."""
+    base = capabilities_for_tool(tool_name)
+    if not isinstance(tool_name, str):
+        return base
+
+    action = _action_from_content(tool_name, content)
+    destructive = action in _ACTION_DESTRUCTIVE.get(tool_name, ())
+    if tool_name not in _PRIVATE_ACTION_READS:
+        if not destructive:
+            return base
+        return ToolCapabilities(
+            frozenset(set(base.effects) | {ToolEffect.DESTRUCTIVE}),
+            base.result_integrity,
+            known=base.known,
+        )
+    if action in _PRIVATE_ACTION_READS[tool_name]:
+        return _capabilities(
+            ToolEffect.READ_PRIVATE,
+            result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+        )
+    if action in _PRIVATE_ACTION_WRITES[tool_name]:
+        effects = set(base.effects)
+        if destructive:
+            effects.add(ToolEffect.DESTRUCTIVE)
+        return ToolCapabilities(
+            frozenset(effects),
+            ResultIntegrity.EXTERNAL_UNTRUSTED,
+            known=base.known,
+        )
+
+    return _capabilities(
+        ToolEffect.READ_PRIVATE,
+        ToolEffect.WRITE_PRIVATE,
+        result_integrity=ResultIntegrity.EXTERNAL_UNTRUSTED,
+    )
+
+
+def tool_result_is_successful(result: Any) -> bool:
+    """Return whether a result actually introduced successful tool output."""
+    return bool(
+        isinstance(result, dict)
+        and not result.get("blocked")
+        and not result.get("approval_required")
+        and not result.get("error")
+        and result.get("exit_code") in (None, 0)
+        and result.get("success") is not False
+    )
+
+
+def tool_result_should_arm_gate(
+    tool_name: Any,
+    result: Any,
+    content: Any = None,
+) -> bool:
+    """Return whether a result introduced non-system content to the model.
+
+    A blocked/approval placeholder and a genuinely content-free failure do not
+    change authority. Once a non-system tool returns text or structured data
+    that will be folded into model context, however, failure status cannot make
+    that payload trusted: MCP ``isError`` text, provider exception messages,
+    and HTTP error bodies are all attacker-controlled input surfaces.
+    """
+    if not isinstance(result, dict):
+        return False
+    if result.get("blocked") or result.get("approval_required"):
+        return False
+    # A producer that knows a particular response body came from a remote or
+    # stored source overrides a coarse static SYSTEM default.
+    if result.get("untrusted_content") is True:
+        return True
+    capabilities = capabilities_for_action(tool_name, content)
+    if capabilities.result_integrity is ResultIntegrity.SYSTEM:
+        return False
+    if tool_result_is_successful(result):
+        return True
+    # ``format_tool_result`` serializes every additional structured field, so
+    # a fixed allowlist here would inevitably miss model-visible payloads such
+    # as ``details``, ``events``, or provider-specific response keys. Exclude
+    # only status/policy controls that carry no producer content; any other
+    # non-empty field crosses the same integrity boundary even on failure.
+    non_content_keys = frozenset(
+        {
+            "approval_required",
+            "blocked",
+            "exit_code",
+            "policy",
+            "success",
+            "untrusted_content",
+        }
+    )
+    return any(
+        key not in non_content_keys and value not in (None, "", [], {}, ())
+        for key, value in result.items()
+    )
+
+
+POST_EXTERNAL_BLOCKED_EFFECTS = frozenset(
+    {
+        ToolEffect.READ_PRIVATE,
+        ToolEffect.WRITE_WORKSPACE,
+        ToolEffect.WRITE_PRIVATE,
+        ToolEffect.EXECUTE_CODE,
+        ToolEffect.NETWORK_EGRESS,
+        ToolEffect.EXTERNAL_SIDE_EFFECT,
+        ToolEffect.UI_SIDE_EFFECT,
+        ToolEffect.ADMIN_CHANGE,
+        ToolEffect.DESTRUCTIVE,
+    }
+)
+
+
+@dataclass(frozen=True)
+class ToolGateDecision:
+    allowed: bool
+    reason: str | None = None
+
+
+_EXTERNAL_MESSAGE_SOURCES = frozenset(
+    {
+        "injected research context",
+        "prefetched search context",
+        "research context",
+        "web search results",
+        "youtube transcript",
+    }
+)
+_EXTERNAL_MESSAGE_SOURCE_PREFIXES = ("web page:",)
+
+
+def messages_contain_external_untrusted_context(messages: Iterable[dict]) -> bool:
+    """Detect explicitly labelled external context already present in a run."""
+    for message in messages or ():
+        if not isinstance(message, dict):
+            continue
+        metadata = message.get("metadata")
+        if not isinstance(metadata, dict) or metadata.get("trusted") is not False:
+            continue
+        gate_marker = metadata.get("tool_gate_untrusted")
+        if gate_marker is True:
+            return True
+        if gate_marker is False:
+            # Explicit current-format opt-outs are authoritative.  The source
+            # label heuristics below exist only for older saved wrappers that
+            # predate the marker.
+            continue
+        if metadata.get("provenance_origin") == "external":
+            return True
+        source = metadata.get("source")
+        if not isinstance(source, str):
+            continue
+        normalized_source = source.strip().casefold()
+        if normalized_source in _EXTERNAL_MESSAGE_SOURCES:
+            return True
+        if normalized_source.startswith(_EXTERNAL_MESSAGE_SOURCE_PREFIXES):
+            return True
+    return False
+
+
+@dataclass
+class ToolRunSecurityContext:
+    """Server-owned integrity state for one agent run."""
+
+    external_untrusted_context_seen: bool = False
+    external_sources: list[str] = field(default_factory=list)
+    run_id: str = field(default_factory=lambda: uuid.uuid4().hex)
+
+    def observe_messages(self, messages: Iterable[dict]) -> None:
+        """Promote any server-labelled untrusted prompt context into the gate."""
+        if messages_contain_external_untrusted_context(messages):
+            self.external_untrusted_context_seen = True
+
+    def decision_for(self, tool_name: Any, content: Any = None) -> ToolGateDecision:
+        if not self.external_untrusted_context_seen:
+            return ToolGateDecision(True)
+        capabilities = capabilities_for_action(tool_name, content)
+        blocked_effects = capabilities.effects & POST_EXTERNAL_BLOCKED_EFFECTS
+        if capabilities.known and not blocked_effects:
+            return ToolGateDecision(True)
+        effects = ", ".join(sorted(effect.value for effect in blocked_effects))
+        if not capabilities.known:
+            effects = "unknown/high-impact"
+        return ToolGateDecision(
+            False,
+            (
+                "External untrusted context has already influenced this run. "
+                f"Tool '{tool_name}' requires a separate user-authorized action "
+                f"because it can cause {effects}."
+            ),
+        )
+
+    def observe_tool_result(
+        self,
+        tool_name: Any,
+        result: Any,
+        content: Any = None,
+    ) -> None:
+        if not tool_result_should_arm_gate(tool_name, result, content):
+            return
+        self.external_untrusted_context_seen = True
+        if isinstance(tool_name, str) and tool_name not in self.external_sources:
+            self.external_sources.append(tool_name)
+
+
+def blocked_tool_result(tool_name: Any, reason: str) -> tuple[str, dict]:
+    return (
+        f"{tool_name}: BLOCKED",
+        {
+            "error": reason,
+            "exit_code": 1,
+            "blocked": True,
+            "policy": "external_untrusted_context",
+        },
+    )

src/tool_execution.py

@@ -27,10 +27,24 @@ from src.tool_security import (
     is_public_blocked_tool,
     owner_is_admin_or_single_user,
 )
+from src.tool_capabilities import ToolRunSecurityContext, blocked_tool_result
+from src.tool_approvals import ExactToolApproval
 from src.tool_policy import ToolPolicy
 from src.constants import MAX_OUTPUT_CHARS, MAX_READ_CHARS, MAX_DIFF_LINES, DATA_DIR
 from src.tool_utils import _truncate, get_mcp_manager
 
+
+class _MissingToolSecurityContext:
+    pass
+
+
+class _NoToolSecurityContext:
+    """Explicit sentinel for non-agent callers that have no run provenance."""
+
+
+_MISSING_TOOL_SECURITY_CONTEXT = _MissingToolSecurityContext()
+NO_TOOL_SECURITY_CONTEXT = _NoToolSecurityContext()
+
 # Persistent working directory for agent subprocesses.
 # Resolves to <repo_root>/data, which is the bind-mounted volume in Docker
 # (/app/data) and the local data directory for manual installs.
@@ -554,10 +568,19 @@ async def _document_tool_dispatch(
     content: str,
     session_id: Optional[str] = None,
     owner: Optional[str] = None,
+    document_id: Optional[str] = None,
+    document_version: Optional[int] = None,
+    document_digest: Optional[str] = None,
 ) -> Optional[Dict]:
     """Route a document tool through TOOL_HANDLERS with the right ctx shape."""
     from src.agent_tools import TOOL_HANDLERS
-    ctx = {"session_id": session_id, "owner": owner}
+    ctx = {
+        "session_id": session_id,
+        "owner": owner,
+        "doc_id": document_id,
+        "expected_document_version": document_version,
+        "expected_document_digest": document_digest,
+    }
     if tool in TOOL_HANDLERS:
         return await TOOL_HANDLERS[tool](content, ctx)
     return None
@@ -575,6 +598,12 @@ async def execute_tool_block(
     progress_cb: Optional[Callable[[Dict], Awaitable[None]]] = None,
     workspace: Optional[str] = None,
     tool_policy: Optional[Any] = None,
+    security_context: (
+        ToolRunSecurityContext
+        | _NoToolSecurityContext
+        | _MissingToolSecurityContext
+    ) = _MISSING_TOOL_SECURITY_CONTEXT,
+    exact_approval: Optional[ExactToolApproval] = None,
 ) -> Tuple[str, Dict]:
     """Execute a single tool block. Returns (description, result_dict).
 
@@ -582,6 +611,104 @@ async def execute_tool_block(
     cwd confine to it) for the duration of this call, then delegate. Reset on the
     way out so the binding never leaks to the next tool call.
     """
+    if security_context is _MISSING_TOOL_SECURITY_CONTEXT:
+        raise TypeError(
+            "execute_tool_block requires security_context; pass a "
+            "ToolRunSecurityContext or NO_TOOL_SECURITY_CONTEXT explicitly"
+        )
+    if (
+        not isinstance(security_context, ToolRunSecurityContext)
+        and security_context is not NO_TOOL_SECURITY_CONTEXT
+    ):
+        raise TypeError(
+            "security_context must be a ToolRunSecurityContext or "
+            "NO_TOOL_SECURITY_CONTEXT"
+        )
+
+    approval_claimed = False
+    if exact_approval is not None:
+        if (
+            not isinstance(security_context, ToolRunSecurityContext)
+            or not security_context.external_untrusted_context_seen
+            or not exact_approval.pending.external_untrusted_context_seen
+        ):
+            return (
+                f"{getattr(block, 'tool_type', None)}: BLOCKED",
+                {
+                    "error": "Exact-action approval requires an armed run security context.",
+                    "exit_code": 1,
+                    "blocked": True,
+                    "policy": "exact_tool_approval",
+                },
+            )
+        if (
+            exact_approval.pending.tool_name
+            in {"edit_document", "suggest_document", "update_document"}
+            and (
+                not exact_approval.pending.document_id
+                or exact_approval.pending.document_version is None
+                or not exact_approval.pending.document_digest
+            )
+        ):
+            return (
+                f"{getattr(block, 'tool_type', None)}: BLOCKED",
+                {
+                    "error": (
+                        "The approved document action has no sealed target and "
+                        "cannot be executed."
+                    ),
+                    "exit_code": 1,
+                    "blocked": True,
+                    "policy": "exact_tool_approval",
+                },
+            )
+        sealed_workspace = exact_approval.pending.workspace
+        if sealed_workspace and vet_workspace(sealed_workspace) != sealed_workspace:
+            return (
+                f"{getattr(block, 'tool_type', None)}: BLOCKED",
+                {
+                    "error": (
+                        "The approved workspace is no longer a valid safe "
+                        "directory. Review the action again."
+                    ),
+                    "exit_code": 1,
+                    "blocked": True,
+                    "policy": "exact_tool_approval",
+                },
+            )
+        approval_claimed = exact_approval.claim(
+            owner=owner,
+            session_id=session_id,
+            tool_name=getattr(block, "tool_type", None),
+            content=getattr(block, "content", None),
+            workspace=workspace,
+        )
+        if not approval_claimed:
+            return (
+                f"{getattr(block, 'tool_type', None)}: BLOCKED",
+                {
+                    "error": "The exact-action approval did not match this tool request.",
+                    "exit_code": 1,
+                    "blocked": True,
+                    "policy": "exact_tool_approval",
+                },
+            )
+
+    if isinstance(security_context, ToolRunSecurityContext) and not approval_claimed:
+        decision = security_context.decision_for(
+            getattr(block, "tool_type", None),
+            getattr(block, "content", None),
+        )
+        if not decision.allowed:
+            logger.warning(
+                "External-context policy blocked tool=%r",
+                getattr(block, "tool_type", None),
+            )
+            return blocked_tool_result(
+                getattr(block, "tool_type", None),
+                decision.reason or "Tool blocked by external-context policy.",
+            )
+
     token = _active_workspace.set(workspace or None)
     try:
         output = await _execute_tool_block_impl(
@@ -591,7 +718,28 @@ async def execute_tool_block(
             owner=owner,
             progress_cb=progress_cb,
             tool_policy=tool_policy,
+            approved_document_id=(
+                exact_approval.pending.document_id
+                if approval_claimed
+                else None
+            ),
+            approved_document_version=(
+                exact_approval.pending.document_version
+                if approval_claimed
+                else None
+            ),
+            approved_document_digest=(
+                exact_approval.pending.document_digest
+                if approval_claimed
+                else None
+            ),
         )
+        if isinstance(security_context, ToolRunSecurityContext):
+            security_context.observe_tool_result(
+                getattr(block, "tool_type", None),
+                output[1],
+                getattr(block, "content", None),
+            )
         return output
     finally:
         _active_workspace.reset(token)
@@ -604,6 +752,9 @@ async def _execute_tool_block_impl(
     owner: Optional[str] = None,
     progress_cb: Optional[Callable[[Dict], Awaitable[None]]] = None,
     tool_policy: Optional[Any] = None,
+    approved_document_id: Optional[str] = None,
+    approved_document_version: Optional[int] = None,
+    approved_document_digest: Optional[str] = None,
 ) -> Tuple[str, Dict]:
     """Execute a single tool block. Returns (description, result_dict).
 
@@ -765,7 +916,15 @@ async def _execute_tool_block_impl(
     elif tool in ("create_document", "update_document", "edit_document",
                   "suggest_document", "manage_documents"):
         desc = f"{tool}: {content.split(chr(10))[0][:80]}"
-        result = await _document_tool_dispatch(tool, content, session_id, owner) \
+        result = await _document_tool_dispatch(
+            tool,
+            content,
+            session_id,
+            owner,
+            document_id=approved_document_id,
+            document_version=approved_document_version,
+            document_digest=approved_document_digest,
+        ) \
             or {"error": f"{tool}: execution failed", "exit_code": 1}
         if tool in ("edit_document", "suggest_document") and "title" in (result or {}):
             desc = f"{tool}: {result.get('title', '')}"

src/tools/cookbook.py

@@ -954,7 +954,11 @@ async def _cookbook_kill_session(session_id: str, *, remote_host: str = "",
             resp = await client.post(f"{_INTERNAL_BASE}/api/shell/exec",
                                      json={"command": cmd}, headers=headers)
         if resp.status_code >= 400:
-            return {"error": f"shell/exec returned HTTP {resp.status_code}: {resp.text[:200]}", "exit_code": 1}
+            return {
+                "error": f"shell/exec returned HTTP {resp.status_code}: {resp.text[:200]}",
+                "exit_code": 1,
+                "untrusted_content": True,
+            }
         try:
             data = resp.json()
         except Exception:
@@ -1083,7 +1087,11 @@ async def do_tail_serve_output(content: str, owner: Optional[str] = None) -> Dic
             resp = await client.post(f"{_INTERNAL_BASE}/api/shell/exec",
                                      json={"command": cmd}, headers=headers)
         if resp.status_code >= 400:
-            return {"error": f"shell/exec returned HTTP {resp.status_code}: {resp.text[:200]}", "exit_code": 1}
+            return {
+                "error": f"shell/exec returned HTTP {resp.status_code}: {resp.text[:200]}",
+                "exit_code": 1,
+                "untrusted_content": True,
+            }
         data = resp.json() if resp.content else {}
         output_text = (data.get("stdout") or "").strip()
         stderr_text = (data.get("stderr") or "").strip()

src/tools/research.py

@@ -123,7 +123,11 @@ async def do_trigger_research(content: str, owner: Optional[str] = None) -> Dict
             resp = await client.post(f"{_INTERNAL_BASE}/api/research/start",
                                      json=payload, headers=_internal_headers(owner))
         if resp.status_code >= 400:
-            return {"error": f"research/start returned HTTP {resp.status_code}: {resp.text[:200]}", "exit_code": 1}
+            return {
+                "error": f"research/start returned HTTP {resp.status_code}: {resp.text[:200]}",
+                "exit_code": 1,
+                "untrusted_content": True,
+            }
         data = resp.json()
         sid = data.get("session_id", "?")
         return {

src/tools/system.py

@@ -725,6 +725,7 @@ async def do_app_api(content: str, owner: Optional[str] = None) -> Dict:
                 "status_code": resp.status_code,
                 "body": preview,
                 "exit_code": 1,
+                "untrusted_content": True,
             }
         return {
             "output": f"{method} {path} -> {resp.status_code}\n{preview}",

start-macos.sh

@@ -34,6 +34,10 @@ fi
 # values (APP_PORT / APP_BIND), then built-in defaults.
 PORT="${ODYSSEUS_PORT:-${APP_PORT:-7860}}"   # 7860, not 7000 — macOS AirPlay Receiver holds 7000.
 HOST="${ODYSSEUS_HOST:-${APP_BIND:-127.0.0.1}}" # Set APP_BIND=0.0.0.0 in .env for LAN/Tailscale access.
+# The port only reaches uvicorn as a flag, so export it too: everything that
+# builds a URL for this instance — internal_api_base(), the companion pairing
+# code, the MCP OAuth callback — reads APP_PORT and would otherwise assume 7000.
+export APP_PORT="$PORT"
 PROBE_HOST="$HOST"
 if [ "$PROBE_HOST" = "0.0.0.0" ] || [ "$PROBE_HOST" = "::" ]; then
     PROBE_HOST="127.0.0.1"

static/app.js

@@ -10,9 +10,9 @@ import modelsModule from './js/models.js?v=20260715startupcalm2';
 import ragModule from './js/rag.js';
 import presetsModule from './js/presets.js';
 import searchModule from './js/search.js';
-import chatModule from './js/chat.js?v=20260801fix1';
+import chatModule from './js/chat.js?v=20260815toolapproval4';
 import compareModule from './js/compare/index.js?v=20260723compareicon2';
-import documentModule from './js/document.js?v=20260722emailfastindex1';
+import documentModule from './js/document.js?v=20260815approvalsave1';
 import searchChatModule from './js/search-chat.js';
 import { makeWindowDraggable } from './js/windowDrag.js';
 import {
@@ -22,7 +22,7 @@ import {
   settleSessionHydration
 } from './js/startupShell.js';
 import markdownModule from './js/markdown.js';
-import chatRenderer from './js/chatRenderer.js?v=20260722emailfastindex1';
+import chatRenderer from './js/chatRenderer.js?v=20260815toolapproval4';
 import sessionModule from './js/sessions.js';
 import memoryModule from './js/memory.js?v=20260722memoryloading1';
 import voiceRecorderModule from './js/voiceRecorder.js';
@@ -33,7 +33,7 @@ import tasksModule from './js/tasks.js?v=20260723tasksbulkfeedback1';
 import calendarModule from './js/calendar.js';
 import notesModule from './js/notes.js';
 import adminModule from './js/admin.js?v=20260716openrouter3';
-import settingsModule from './js/settings.js?v=20260722emailfastindex1';
+import settingsModule from './js/settings.js?v=20260815approvalsave1';
 // Eagerly bind unified minimize/restore behavior across all tool modals.
 import './js/modalManager.js?v=20260723compareicon2';
 // Desktop window tiling — drag a modal near an edge/corner to snap.
@@ -50,6 +50,7 @@ import * as researchPanelModule from './js/research/panel.js?v=20260630researcht
 import ttsModule from './js/tts-ai.js';
 import spinnerModule from './js/spinner.js';
 import { initKeyboardShortcuts } from './js/keyboard-shortcuts.js';
+import { getSettings } from './js/appConfig.js';
 import { initSidebarLayout, syncRailSide } from './js/sidebar-layout.js?v=20260715startupclean';
 import { initSectionCollapse, initSectionDrag } from './js/section-management.js';
 
@@ -1518,13 +1519,11 @@ function initializeEventListeners() {
     })
     .catch(() => {});
 
-  // Hide Gallery when image generation is disabled in settings
-  const _prefetchedSettings = sessionStorage.getItem('ody-prefetch-settings');
-  sessionStorage.removeItem('ody-prefetch-settings');
-  window._initSettingsReady = (_prefetchedSettings
-    ? Promise.resolve(JSON.parse(_prefetchedSettings))
-    : fetch(`${API_BASE}/api/auth/settings`, { credentials: 'same-origin' }).then(r => r.json())
-  ).then(settings => {
+  // Hide Gallery when image generation is disabled in settings.
+  // getSettings() consumes the login prefetch itself, so every other module
+  // that asks for settings this load gets the same snapshot without a request.
+  window._initSettingsReady = getSettings()
+    .then(settings => {
       // NOTE: image_gen_enabled only governs *generating* images in chat — the
       // tool is blocked server-side (chat_routes / agent_loop). The Gallery
       // holds uploads and past images too, so it stays visible regardless;
@@ -3705,7 +3704,7 @@ function startOdysseusApp() {
   modelsModule.init(API_BASE);
   ragModule.init(API_BASE);
   presetsModule.init(API_BASE);
-  searchModule.init(API_BASE);
+  searchModule.init();
   chatModule.init(API_BASE);
   chatModule.initListeners();
   groupModule.init(API_BASE);

static/index.html

@@ -231,23 +231,11 @@
       }
     }
   </style>
-  <!-- KaTeX CSS is loaded with media="print" so it doesn't block render,
-       then flipped to "all" via JS after load. Mermaid init runs once the
-       library finishes loading. Both hooks are wired via addEventListener
-       below (inline onload= attrs are blocked by CSP script-src-attr). -->
-  <link id="katex-css" rel="stylesheet" href="https://cdn.jsdelivr.net/npm/katex@0.16.22/dist/katex.min.css" media="print">
-  <script async src="https://cdn.jsdelivr.net/npm/katex@0.16.22/dist/katex.min.js"></script>
-  <script id="mermaid-script" async src="https://cdn.jsdelivr.net/npm/mermaid@11/dist/mermaid.min.js"></script>
-  <script nonce="{{CSP_NONCE}}">
-    (function(){
-      var k = document.getElementById('katex-css');
-      if (k) k.addEventListener('load', function(){ k.media = 'all'; }, { once: true });
-      var m = document.getElementById('mermaid-script');
-      if (m) m.addEventListener('load', function(){
-        if (window.odysseusInitMermaid) window.odysseusInitMermaid();
-      }, { once: true });
-    })();
-  </script>
+  <!-- KaTeX and Mermaid are vendored in /static/lib and pulled in by
+       static/js/markdown.js the first time a page actually renders math or a
+       ```mermaid fence. They used to load here from cdn.jsdelivr.net on every
+       page load, which cost ~985 KB on the wire, broke offline installs, and
+       announced every session to a third party. -->
   <!-- Preload the two faces first paint actually uses: Fira Code 400 and 600,
        the app font and the weight the sidebar and header text render at. They
        are declared in style.css, so without a hint they are only discovered
@@ -258,8 +246,8 @@
   <link rel="preload" as="font" type="font/woff2" crossorigin href="/static/fonts/FiraCode-Regular.woff2">
   <link rel="preload" as="font" type="font/woff2" crossorigin href="/static/fonts/FiraCode-SemiBold.woff2">
   <link rel="stylesheet" href="/static/style.css?v=20260808startupshell1">
-  <link rel="modulepreload" href="/static/app.js?v=20260808startupshell1">
-  <link rel="modulepreload" href="/static/js/chat.js?v=20260801fix1">
+  <link rel="modulepreload" href="/static/app.js?v=20260815toolapproval4">
+  <link rel="modulepreload" href="/static/js/chat.js?v=20260815toolapproval4">
   <link rel="modulepreload" href="/static/js/ui.js">
   <link rel="modulepreload" href="/static/js/sessions.js">
   <link rel="modulepreload" href="/static/js/markdown.js">
@@ -1421,6 +1409,55 @@
       </div>
       <div class="settings-layout">
         <div class="settings-sidebar">
+          <button
+            type="button"
+            class="settings-sidebar-toggle"
+            id="settings-sidebar-toggle"
+            aria-label="Collapse settings navigation"
+            title="Collapse settings navigation"
+          >
+            <svg class="settings-sidebar-toggle-collapse" width="12" height="12" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2.3" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true">
+              <polyline points="15 18 9 12 15 6"></polyline>
+            </svg>
+            <svg class="settings-sidebar-toggle-expand" width="12" height="12" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2.3" stroke-linecap="round" stroke-linejoin="round" aria-hidden="true">
+              <polyline points="9 18 15 12 9 6"></polyline>
+            </svg>
+          </button>
+
+          <div
+            class="settings-sidebar-resize-handle"
+            id="settings-sidebar-resize-handle"
+            role="separator"
+            aria-orientation="vertical"
+            aria-label="Resize settings navigation"
+            aria-valuemin="150"
+            aria-valuemax="340"
+            aria-valuenow="220"
+            tabindex="0"
+          ></div>
+
+          <div class="settings-sidebar-content">
+          <div class="settings-nav-search-wrap">
+            <svg class="settings-nav-search-icon" width="13" height="13" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" aria-hidden="true">
+              <circle cx="11" cy="11" r="7"></circle>
+              <path d="M20 20l-4-4"></path>
+            </svg>
+            <input
+              type="search"
+              id="settings-nav-search"
+              class="settings-nav-search"
+              placeholder="Find settings…"
+              autocomplete="off"
+              aria-label="Find settings"
+              aria-controls="settings-nav-search-results"
+            />
+            <div
+              id="settings-nav-search-results"
+              class="settings-nav-search-results hidden"
+              role="listbox"
+              aria-label="Settings search results"
+            ></div>
+          </div>
           <!-- Section 1: AI plumbing (Add Models → AI Defaults → Search) -->
           <button class="settings-nav-item active" data-settings-tab="services">
             <svg width="15" height="15" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2.6" stroke-linecap="round" stroke-linejoin="round"><path d="M12 5v14"/><path d="M5 12h14"/></svg>
@@ -1481,9 +1518,10 @@
             <span>Users</span>
           </button>
           <button class="settings-nav-item admin-only" data-settings-tab="system">
-            <svg width="15" height="15" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><circle cx="12" cy="12" r="3"/><path d="M19.4 15a1.65 1.65 0 0 0 .33 1.82l.06.06a2 2 0 0 1-2.83 2.83l-.06-.06a1.65 1.65 0 0 0-1.82-.33 1.65 1.65 0 0 0-1 1.51V21a2 2 0 0 1-4 0v-.09A1.65 1.65 0 0 0 9 19.4a1.65 1.65 0 0 0-1.82.33l-.06.06a2 2 0 0 1-2.83-2.83l.06-.06A1.65 1.65 0 0 0 4.68 15a1.65 1.65 0 0 0-1.51-1H3a2 2 0 0 1 0-4h.09A1.65 1.65 0 0 0 4.6 9a1.65 1.65 0 0 0-.33-1.82l-.06-.06a2 2 0 0 1 2.83-2.83l.06.06A1.65 1.65 0 0 0 9 4.68a1.65 1.65 0 0 0 1-1.51V3a2 2 0 0 1 4 0v.09a1.65 1.65 0 0 0 1 1.51 1.65 1.65 0 0 0 1.82-.33l.06-.06a2 2 0 0 1 2.83 2.83l-.06.06A1.65 1.65 0 0 0 19.4 9a1.65 1.65 0 0 0 1.51 1H21a2 2 0 0 1 0 4h-.09a1.65 1.65 0 0 0-1.51 1z"/></svg>
+            <svg width="15" height="15" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><circle cx="12" cy="12" r="3"/><path d="M19.4 15a1.65 1.65 0 0 0 .33 1.82l.06.06a2 2 0 0 1-2.83 2.83l-.06-.06a1.65 1.65 0 0 0-1.82-.33 1.65 1.65 0 0 0-1 1.51V21a2 2 0 0 1-4 0v-.09A1.65 1.65 0 0 0 9 19.4a1.65 1.65 0 0 0-1.82.33l-.06.06a2 2 0 0 1-2.83-2.83l.06-.06A1.65 1.65 0 0 0 4.68 15a1.65 1.65 0 0 0-1.51-1H3a2 2 0 0 1 0-4h.09A1.65 1.65 0 0 0 4.6 9a1.65 1.65 0 0 0-.33-1.82l-.06-.06a2 2 0 0 1 2.83-2.83l.06.06A1.65 1.65 0 0 0 9 4.68a1.65 1.65 0 0 0 1-1.51V3a2 2 0 0 1 4 0v-.09a1.65 1.65 0 0 0 1 1.51 1.65 1.65 0 0 0 1.82-.33l.06-.06a2 2 0 0 1 2.83 2.83l-.06.06A1.65 1.65 0 0 0 19.4 9a1.65 1.65 0 0 0 1.51 1H21a2 2 0 0 1 0 4h-.09A1.65 1.65 0 0 0 19.4 15z"/></svg>
             <span>System</span>
           </button>
+          </div>
         </div>
         <div class="settings-panels">
 
@@ -1504,13 +1542,6 @@
                 <span class="adm-model-logo" id="set-defaultModelSelect-logo" style="display:inline-flex;align-items:center;justify-content:center;width:18px;height:18px;flex-shrink:0;opacity:0.9;color:var(--fg);"></span>
                 <select id="set-defaultModelSelect" class="settings-select"></select>
               </div>
-              <div class="settings-row" style="align-items:flex-start;" hidden>
-                <label class="settings-label" style="margin-top:6px;">Fallbacks</label>
-                <div style="flex:1;display:flex;flex-direction:column;gap:6px;">
-                  <div id="set-defaultFallbacks" class="settings-fallbacks"></div>
-                  <button type="button" class="settings-fallback-add" id="set-defaultAddFallback" title="Add a model to try if the one above fails">+ Add fallback</button>
-                </div>
-              </div>
               <div id="set-defaultChatMsg" style="font-size:11px;color:color-mix(in srgb, var(--fg) 45%, transparent);"></div>
             </div>
           </div>
@@ -2539,20 +2570,20 @@
 <script type="module" src="/static/js/search.js"></script>
 <script type="module" src="/static/js/spinner.js"></script>
 <script type="module" src="/static/js/tts-ai.js"></script>
-<script type="module" src="/static/js/document.js?v=20260722emailfastindex1"></script>
+<script type="module" src="/static/js/document.js?v=20260815approvalsave1"></script>
 <script type="module" src="/static/js/gallery.js?v=20260708match1"></script>
-<script type="module" src="/static/js/chatRenderer.js?v=20260722emailfastindex1"></script>
+<script type="module" src="/static/js/chatRenderer.js?v=20260815toolapproval4"></script>
 <script type="module" src="/static/js/codeRunner.js"></script>
-<script type="module" src="/static/js/chatStream.js?v=20260722emailfastindex1"></script>
-  <script type="module" src="/static/js/chat.js?v=20260801fix1"></script>
+<script type="module" src="/static/js/chatStream.js?v=20260815approvalsave1"></script>
+  <script type="module" src="/static/js/chat.js?v=20260815toolapproval4"></script>
 <script type="module" src="/static/js/cookbook.js"></script>
 <script src="/static/js/cookbookSchedule.js"></script>
 <script type="module" src="/static/js/search-chat.js"></script>
 <script type="module" src="/static/js/theme.js"></script>
 <script type="module" src="/static/js/censor.js"></script>
-<script type="module" src="/static/js/settings.js?v=20260723compareicon1"></script>
+<script type="module" src="/static/js/settings.js?v=20260815approvalsave1"></script>
 <script type="module" src="/static/js/assistant.js"></script>
-<script type="module" src="/static/app.js?v=20260808startupshell1"></script>  <!-- app.js must be LAST -->
+<script type="module" src="/static/app.js?v=20260815toolapproval4"></script>  <!-- app.js must be LAST -->
   <script type="module" src="/static/js/init.js?v=20260715freshroot3"></script>
 <script type="module" src="/static/js/a11y.js"></script>
 <script nonce="{{CSP_NONCE}}">if('serviceWorker' in navigator){navigator.serviceWorker.register('/static/sw.js').catch(()=>{});}</script>

static/js/admin.js

@@ -6,6 +6,7 @@ import settingsModule from './settings.js';
 import { providerLogo, providerLogoFromUrl } from './providers.js';
 import { sortModelObjects } from './modelSort.js';
 import { PROVIDER_DEVICE_FLOWS, formatDeviceFlowError, runProviderDeviceFlow } from './providerDeviceFlow.js';
+import { getSettings, getTools, invalidateSettings, invalidateTools } from './appConfig.js';
 
 let initialized = false;
 let modalEl = null;
@@ -345,8 +346,7 @@ function initSignupToggle() {
 
 function initShareDefaultsToggle() {
   const toggle = el('adm-shareDefaultsToggle');
-  fetch('/api/auth/settings', { credentials: 'same-origin' })
-    .then(r => r.json())
+  getSettings()
     .then(d => { toggle.checked = !!d.share_defaults_with_users; })
     .catch(e => console.warn('Settings fetch failed:', e));
   toggle.addEventListener('change', async () => {
@@ -361,6 +361,9 @@ function initShareDefaultsToggle() {
       toggle.checked = !!data.share_defaults_with_users;
     } catch (e) {
       toggle.checked = !toggle.checked;
+    } finally {
+      // Drop the shared snapshot: it still says what this toggle used to be.
+      invalidateSettings();
     }
   });
 }
@@ -1893,8 +1896,16 @@ async function loadBuiltinTools() {
   const list = el('adm-builtin-tools-list');
   if (!list) return;
   try {
-    const res = await fetch('/api/tools', { credentials: 'same-origin' });
-    const data = await res.json();
+    // This panel is an editor, and its save posts the whole disabled list
+    // rebuilt from the checkboxes below. So it has to render authoritative
+    // state: a snapshot that went stale out of band (the manage_settings tool,
+    // another tab) would be re-posted wholesale on the next unrelated toggle
+    // and would silently undo the newer state. refreshAll() calls this on every
+    // panel open, so drop the shared entry and refill it. The startup read that
+    // chatRenderer.js shares is unaffected; this panel just never edits a cache,
+    // which is the same rule the settings panel follows by reading directly.
+    invalidateTools();
+    const data = await getTools();
     const tools = data.tools || [];
     if (!tools.length) { list.innerHTML = '<div class="admin-empty">No tools found</div>'; return; }
 
@@ -1968,17 +1979,50 @@ async function loadBuiltinTools() {
       });
     });
 
-    // Helper: save disabled tools + update counters
-    async function _saveToolState() {
-      const allChecks = list.querySelectorAll('input[data-tool-id]');
-      const disabled = [];
-      allChecks.forEach(c => { if (!c.checked) disabled.push(c.dataset.toolId); });
-      await fetch('/api/tools', {
-        method: 'POST',
-        headers: { 'Content-Type': 'application/json' },
-        body: JSON.stringify({ disabled }),
-        credentials: 'same-origin',
-      });
+    // Merge only the user's intended changes onto authoritative server state.
+    // /api/tools replaces the full disabled list, so rebuilding it from this
+    // panel's DOM can undo a change made by another tab or manage_settings
+    // after the panel was opened.
+    async function _saveToolState(changes) {
+      invalidateTools();
+      const latest = await getTools();
+      const state = new Map(
+        (latest.tools || []).map(t => [t.id, !!t.enabled])
+      );
+
+      for (const change of changes) {
+        if (state.has(change.id)) {
+          state.set(change.id, !!change.enabled);
+        }
+      }
+
+      const disabled = Array.from(state.entries())
+        .filter(([, enabled]) => !enabled)
+        .map(([id]) => id);
+
+      try {
+        const res = await fetch('/api/tools', {
+          method: 'POST',
+          headers: { 'Content-Type': 'application/json' },
+          body: JSON.stringify({ disabled }),
+          credentials: 'same-origin',
+        });
+        if (!res.ok) throw new Error(`Failed to update tools (${res.status})`);
+
+        // Bring the still-open editor forward to the same merged snapshot so an
+        // out-of-band change is visible instead of leaving stale checkboxes.
+        list.querySelectorAll('input[data-tool-id]').forEach(c => {
+          if (state.has(c.dataset.toolId)) {
+            c.checked = state.get(c.dataset.toolId);
+          }
+        });
+        list.querySelectorAll('.admin-tool-category').forEach(_updateCatCounter);
+      } finally {
+        // This route persists disabled_tools into the settings store
+        // (routes/model_routes.py), so both snapshots are now stale.
+        invalidateTools();
+        invalidateSettings();
+      }
     }
     function _updateCatCounter(catEl) {
       if (!catEl) return;
@@ -1993,7 +2037,9 @@ async function loadBuiltinTools() {
     // Wire individual tool toggles
     list.querySelectorAll('input[data-tool-id]').forEach(chk => {
       chk.addEventListener('change', async () => {
-        await _saveToolState();
+        await _saveToolState([
+          { id: chk.dataset.toolId, enabled: chk.checked },
+        ]);
         _updateCatCounter(chk.closest('.admin-tool-category'));
       });
     });
@@ -2004,8 +2050,10 @@ async function loadBuiltinTools() {
         const catEl = chk.closest('.admin-tool-category');
         if (!catEl) return;
         const checked = chk.checked;
+        const changes = Array.from(catEl.querySelectorAll('input[data-tool-id]'))
+          .map(c => ({ id: c.dataset.toolId, enabled: checked }));
         catEl.querySelectorAll('input[data-tool-id]').forEach(c => { c.checked = checked; });
-        await _saveToolState();
+        await _saveToolState(changes);
         _updateCatCounter(catEl);
       });
     });

static/js/appConfig.js

@@ -0,0 +1,86 @@
+// static/js/appConfig.js
+//
+// One shared, invalidatable cache for the two config endpoints that every
+// module wants at startup.
+//
+// Before this, /api/auth/settings was fetched independently by six modules and
+// /api/tools by three, none of them aware of the others — 4 and 3 requests on a
+// single cold load. Worse than the requests: each caller could observe a
+// different snapshot of the same object, and chatRenderer.js is imported under
+// three different ?v= query strings, so it is three separate module instances
+// each issuing its own /api/tools fetch. Caching here fixes both, because the
+// cache lives in one module every instance imports by the same specifier.
+//
+// URLs are bare paths on purpose. The callers that used `${API_BASE}/api/...`
+// resolved to the identical URL — API_BASE is `window.location.origin`
+// (app.js) — so nothing about the request changes for them.
+//
+// WRITERS MUST INVALIDATE. Anything that POSTs /api/auth/settings calls
+// invalidateSettings(); anything that POSTs /api/tools calls invalidateTools()
+// *and* invalidateSettings(), because that route persists `disabled_tools`
+// into the same settings store (routes/model_routes.py). Miss one and the UI
+// serves a stale settings object for the rest of the session, which is worse
+// than the duplicate fetches this replaces.
+//
+// The resolved object is shared by reference, so treat it as read-only: copy
+// before mutating (`{ ...await getSettings() }`).
+
+// Written by login.html immediately before it redirects to '/', so the first
+// load after a login can skip the request entirely. Consumed once per page
+// load, by whichever module asks for settings first.
+const PREFETCH_KEY = 'ody-prefetch-settings';
+
+const _URLS = { settings: '/api/auth/settings', tools: '/api/tools' };
+const _cache = { settings: null, tools: null };
+
+function _readPrefetchedSettings() {
+  try {
+    const raw = sessionStorage.getItem(PREFETCH_KEY);
+    if (!raw) return null;
+    sessionStorage.removeItem(PREFETCH_KEY);
+    return JSON.parse(raw);
+  } catch (_) {
+    return null;
+  }
+}
+
+// A rejected promise must not stay in the slot. Plain `??=` memoisation would
+// keep it, so one transient blip during boot would leave keybinds, TTS and the
+// search provider on their defaults for the whole session with no retry. Clear
+// the slot on failure — unless a later invalidate/refetch already replaced it —
+// and rethrow, so every caller's existing .catch() still runs exactly as before.
+function _get(key) {
+  if (_cache[key]) return _cache[key];
+  const pending = fetch(_URLS[key], { credentials: 'same-origin' })
+    .then(r => r.json())
+    .catch(err => {
+      if (_cache[key] === pending) _cache[key] = null;
+      throw err;
+    });
+  _cache[key] = pending;
+  return pending;
+}
+
+/** GET /api/auth/settings, once per page load (or once per invalidation). */
+export function getSettings() {
+  if (!_cache.settings) {
+    const prefetched = _readPrefetchedSettings();
+    if (prefetched) _cache.settings = Promise.resolve(prefetched);
+  }
+  return _get('settings');
+}
+
+/** GET /api/tools, once per page load (or once per invalidation). */
+export function getTools() {
+  return _get('tools');
+}
+
+/** Call after any write that can change settings. */
+export function invalidateSettings() {
+  _cache.settings = null;
+}
+
+/** Call after any write that can change the tool enable/disable state. */
+export function invalidateTools() {
+  _cache.tools = null;
+}

static/js/chatModelProvenance.js

@@ -0,0 +1,104 @@
+/** Select and update the response holder for a route-provenance event. */
+export function applyModelRouteEventState(event, holder, roundHolder, defaultModel = '') {
+  const target = event && event.round && roundHolder ? roundHolder : holder;
+  if (!target) return null;
+
+  target._requestedModel = (
+    event.requested_model
+    || event.selected_model
+    || target._requestedModel
+    || defaultModel
+  );
+  target._actualModel = (
+    event.model
+    || event.answered_by
+    || target._actualModel
+    || target._requestedModel
+  );
+  const hasEndpointRoute = Boolean(
+    event.requested_endpoint_id
+    || event.selected_endpoint_id
+    || event.endpoint_id
+    || event.answered_by_endpoint_id
+    || event.requested_endpoint_label
+    || event.selected_endpoint_label
+    || event.endpoint_label
+    || event.answered_by_endpoint_label
+    || target._requestedEndpointLabel
+  );
+  if (hasEndpointRoute) {
+    target._requestedEndpointId = (
+      event.requested_endpoint_id
+      || event.selected_endpoint_id
+      || target._requestedEndpointId
+      || null
+    );
+    target._requestedEndpointLabel = (
+      event.requested_endpoint_label
+      || event.selected_endpoint_label
+      || target._requestedEndpointLabel
+      || 'Selected route'
+    );
+    target._actualEndpointId = (
+      event.endpoint_id
+      || event.answered_by_endpoint_id
+      || target._actualEndpointId
+      || target._requestedEndpointId
+      || null
+    );
+    target._actualEndpointLabel = (
+      event.endpoint_label
+      || event.answered_by_endpoint_label
+      || target._actualEndpointLabel
+      || target._requestedEndpointLabel
+    );
+  }
+  return target;
+}
+
+/** Copy the active route into the bubble created for the next Agent round. */
+export function inheritModelRouteState(holder, roundHolder, target, defaultModel = '') {
+  if (!target) return null;
+  const source = roundHolder || holder;
+  target._requestedModel = source?._requestedModel || defaultModel;
+  target._actualModel = source?._actualModel || target._requestedModel;
+  if (source?._requestedEndpointLabel || source?._actualEndpointLabel) {
+    target._requestedEndpointId = source?._requestedEndpointId || null;
+    target._requestedEndpointLabel = source?._requestedEndpointLabel || 'Selected route';
+    target._actualEndpointId = source?._actualEndpointId || target._requestedEndpointId;
+    target._actualEndpointLabel = source?._actualEndpointLabel || target._requestedEndpointLabel;
+  }
+  return target;
+}
+
+/** Apply final/metrics provenance to the active round, not the first bubble. */
+export function applyModelMetricsState(metrics, holder, roundHolder, defaultModel = '') {
+  const target = roundHolder || holder;
+  if (!target || !metrics) return target || null;
+  const roundModels = Array.isArray(metrics.round_models) ? metrics.round_models : [];
+  const roundModel = roundHolder && roundModels.length
+    ? roundModels[roundModels.length - 1]
+    : null;
+  target._requestedModel = metrics.requested_model || target._requestedModel || defaultModel;
+  target._actualModel = roundModel || metrics.model || target._actualModel || target._requestedModel;
+  const roundEndpointIds = Array.isArray(metrics.round_endpoint_ids) ? metrics.round_endpoint_ids : [];
+  const roundEndpointLabels = Array.isArray(metrics.round_endpoint_labels) ? metrics.round_endpoint_labels : [];
+  if (
+    metrics.requested_endpoint_label
+    || metrics.endpoint_label
+    || roundEndpointLabels.length
+    || target._requestedEndpointLabel
+  ) {
+    target._requestedEndpointId = metrics.requested_endpoint_id || target._requestedEndpointId || null;
+    target._requestedEndpointLabel = metrics.requested_endpoint_label || target._requestedEndpointLabel || 'Selected route';
+    const hasRoundEndpointId = Boolean(roundHolder && roundEndpointIds.length);
+    const hasRoundEndpointLabel = Boolean(roundHolder && roundEndpointLabels.length);
+    target._actualEndpointId = hasRoundEndpointId
+      ? roundEndpointIds[roundEndpointIds.length - 1]
+      : (metrics.endpoint_id || target._actualEndpointId || target._requestedEndpointId);
+    target._actualEndpointLabel = hasRoundEndpointLabel
+      ? roundEndpointLabels[roundEndpointLabels.length - 1]
+      : (metrics.endpoint_label || target._actualEndpointLabel || target._requestedEndpointLabel);
+  }
+  return target;
+}

static/js/chatRenderer.js

@@ -9,7 +9,9 @@ import { providerLogo, providerLabel } from './providers.js';
 import settingsModule from './settings.js';
 import spinnerModule from './spinner.js';
 import { bindMenuDismiss } from './escMenuStack.js';
+import { loadPanel } from './panels.js';
 import { matchModelKey } from './model/matchKey.js';
+import { getTools } from './appConfig.js';
 
 const SEARCH_ICON = '<svg viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2"><circle cx="11" cy="11" r="8"/><path d="M21 21l-4.35-4.35"/></svg>';
 const REPORT_ICON = '<svg width="14" height="14" viewBox="0 0 24 24" fill="none" stroke="currentColor" stroke-width="2" stroke-linecap="round" stroke-linejoin="round"><path d="M14 2H6a2 2 0 0 0-2 2v16a2 2 0 0 0 2 2h12a2 2 0 0 0 2-2V8z"/><polyline points="14 2 14 8 20 8"/><line x1="16" y1="13" x2="8" y2="13"/><line x1="16" y1="17" x2="8" y2="17"/><line x1="10" y1="9" x2="8" y2="9"/></svg>';
@@ -445,8 +447,12 @@ function stripExecutedFence(match, tag, inline, body) {
 
 async function loadExecFenceRegex() {
   try {
-    const res = await fetch('/api/tools', { credentials: 'same-origin' });
-    const data = await res.json();
+    // Shared with admin.js, and — more to the point — with the other copies of
+    // this module: chatRenderer.js is imported under three different ?v= query
+    // strings, so it is instantiated three times per load and used to issue
+    // three identical /api/tools requests. appConfig.js is imported by one
+    // specifier from all of them, so they now share a single fetch.
+    const data = await getTools();
     const tags = (data.tools || [])
       .map((t) => t.id)
       .filter((id) => id && !EXEC_FENCE_NON_TOOL.has(id));
@@ -615,10 +621,36 @@ export function sameModelName(left, right) {
     || shortModel(a).toLowerCase() === shortModel(b).toLowerCase();
 }
 
-export function modelRouteLabel(requestedModel, actualModel) {
+function shortEndpointLabel(label) {
+  const value = modelValue(label);
+  if (!value) return '';
+  return value.length > 18 ? value.slice(0, 17) + '…' : value;
+}
+
+export function modelRouteLabel(
+  requestedModel,
+  actualModel,
+  requestedEndpointLabel = '',
+  actualEndpointLabel = '',
+  requestedEndpointId = '',
+  actualEndpointId = '',
+) {
   const requested = modelValue(requestedModel);
   const actual = modelValue(actualModel) || requested;
-  if (!requested || sameModelName(requested, actual)) return shortModel(actual || requested);
+  const requestedRoute = modelValue(requestedEndpointId || requestedEndpointLabel);
+  const actualRoute = modelValue(actualEndpointId || actualEndpointLabel);
+  const routeChanged = Boolean(
+    actualRoute
+    && requestedRoute
+    && actualRoute !== requestedRoute
+  );
+  if (!requested || sameModelName(requested, actual)) {
+    const model = shortModel(actual || requested);
+    if (!routeChanged) return model;
+    const from = shortEndpointLabel(requestedEndpointLabel || 'Selected route');
+    const to = shortEndpointLabel(actualEndpointLabel || actualEndpointId);
+    return model + ' (' + from + ' -> ' + to + ')';
+  }
   return shortModel(requested) + ' -> ' + shortModel(actual);
 }
 
@@ -629,10 +661,24 @@ export function replyModelPair(modelName, metadata) {
   if (actualFromMeta || requestedFromMeta) {
     const actual = actualFromMeta || requestedFromMeta || modelValue(modelName);
     const requested = requestedFromMeta || actual;
-    return { requestedModel: requested, actualModel: actual };
+    return {
+      requestedModel: requested,
+      actualModel: actual,
+      requestedEndpointId: meta.requested_endpoint_id || null,
+      requestedEndpointLabel: meta.requested_endpoint_label || 'Selected route',
+      actualEndpointId: meta.endpoint_id || null,
+      actualEndpointLabel: meta.endpoint_label || meta.requested_endpoint_label || 'Selected route',
+    };
   }
   const fallback = modelValue(modelName);
-  return { requestedModel: fallback, actualModel: fallback };
+  return {
+    requestedModel: fallback,
+    actualModel: fallback,
+    requestedEndpointId: null,
+    requestedEndpointLabel: 'Selected route',
+    actualEndpointId: null,
+    actualEndpointLabel: 'Selected route',
+  };
 }
 
 /**
@@ -824,12 +870,50 @@ export function isCostTrackedEndpoint(url) {
 }
 
 /** Cost for the current turn, returning null for non-billable endpoints. */
-function _billableCost(model, inputTokens, outputTokens) {
-  const url = _currentEndpointUrl();
-  if (!isCostTrackedEndpoint(url)) return null;
+function _billableCost(model, inputTokens, outputTokens, endpointCostTracked, selectedEndpointUrl) {
+  // Foreground fallback can answer on a different endpoint than the session's
+  // selected route. Prefer the backend's non-secret actual-route
+  // classification; retain the selected-endpoint check for older history.
+  if (endpointCostTracked === false) return null;
+  const selectedUrl = selectedEndpointUrl === undefined
+    ? _currentEndpointUrl()
+    : selectedEndpointUrl;
+  if (endpointCostTracked !== true && !isCostTrackedEndpoint(selectedUrl)) {
+    return null;
+  }
   return getModelCost(model, inputTokens, outputTokens);
 }
 
+/** Sum cost using the route/model that produced each Agent round. */
+function _metricsBillableCost(metrics, model, inputTokens, outputTokens, selectedEndpointUrl) {
+  const buckets = Array.isArray(metrics.usage_buckets) ? metrics.usage_buckets : [];
+  if (!buckets.length) {
+    return _billableCost(
+      model,
+      inputTokens,
+      outputTokens,
+      metrics.endpoint_cost_tracked,
+      selectedEndpointUrl,
+    );
+  }
+  let total = 0;
+  let hasPricedUsage = false;
+  for (const bucket of buckets) {
+    if (!bucket || typeof bucket !== 'object') continue;
+    const bucketCost = _billableCost(
+      bucket.model || model,
+      Number(bucket.input_tokens) || 0,
+      Number(bucket.output_tokens) || 0,
+      bucket.endpoint_cost_tracked,
+      selectedEndpointUrl,
+    );
+    if (bucketCost === null) continue;
+    total += bucketCost;
+    hasPricedUsage = true;
+  }
+  return hasPricedUsage ? total : null;
+}
+
 export function getImageCost(model, quality, size) {
   if (!model) return null;
   const m = model.toLowerCase();
@@ -844,6 +928,9 @@ export function getImageCost(model, quality, size) {
 
 /* ── Session cost helpers ─────────────────────────────────────────── */
 const _COST_KEY = 'ody-session-cost';
+const _COST_RUNS_KEY = 'ody-session-cost-runs';
+const _MAX_COST_RUNS_PER_SESSION = 256;
+const _COST_LEDGER_LOCK = 'odysseus-session-cost-ledger';
 
 /** Return the accumulated cost for the current (or given) session. */
 export function getSessionCost(sessionId) {
@@ -851,7 +938,14 @@ export function getSessionCost(sessionId) {
   if (!sid) return 0;
   try {
     const costs = JSON.parse(localStorage.getItem(_COST_KEY) || '{}');
-    return costs[sid] || 0;
+    const runCosts = JSON.parse(localStorage.getItem(_COST_RUNS_KEY) || '{}');
+    const recordedRuns = runCosts[sid] && typeof runCosts[sid] === 'object'
+      ? Object.values(runCosts[sid])
+      : [];
+    return (costs[sid] || 0) + recordedRuns.reduce(
+      (total, value) => total + (Number(value) || 0),
+      0,
+    );
   } catch (_e) { return 0; }
 }
 
@@ -863,6 +957,9 @@ export function resetSessionCost(sessionId) {
     const costs = JSON.parse(localStorage.getItem(_COST_KEY) || '{}');
     delete costs[sid];
     localStorage.setItem(_COST_KEY, JSON.stringify(costs));
+    const runCosts = JSON.parse(localStorage.getItem(_COST_RUNS_KEY) || '{}');
+    delete runCosts[sid];
+    localStorage.setItem(_COST_RUNS_KEY, JSON.stringify(runCosts));
   } catch (_e) { /* ignore */ }
   updateSessionCostUI();
 }
@@ -871,21 +968,8 @@ export function resetSessionCost(sessionId) {
 export function updateSessionCostUI() {
   const el = document.getElementById('session-cost-display');
   if (!el) return;
-  // Non-billable endpoint? Hide the badge and clear stale cost that a previous
-  // cloud-rate calculation may have left in localStorage for this session.
-  const _url = _currentEndpointUrl();
-  if (!isCostTrackedEndpoint(_url)) {
-    const sid = window.sessionModule && window.sessionModule.getCurrentSessionId();
-    if (sid && getSessionCost(sid) > 0) {
-      try {
-        const costs = JSON.parse(localStorage.getItem(_COST_KEY) || '{}');
-        delete costs[sid];
-        localStorage.setItem(_COST_KEY, JSON.stringify(costs));
-      } catch (_e) { /* ignore */ }
-    }
-    el.style.display = 'none';
-    return;
-  }
+  // The ledger records billable work already performed in this session. A
+  // selected local endpoint does not erase cost from a paid fallback route.
   const cost = getSessionCost();
   if (cost > 0) {
     el.textContent = '$' + (cost < 0.01 ? cost.toFixed(4) : cost < 1 ? cost.toFixed(3) : cost.toFixed(2));
@@ -895,6 +979,94 @@ export function updateSessionCostUI() {
   }
 }
 
+/** Record one metrics payload in a session ledger at most once. */
+export function recordSessionMetricsCost(metrics, sessionId, selectedEndpointUrl) {
+  if (!metrics || typeof metrics !== 'object') return null;
+  const cost = _metricsBillableCost(
+    metrics,
+    metrics.model || 'Unknown',
+    metrics.input_tokens || 0,
+    metrics.output_tokens || 0,
+    selectedEndpointUrl,
+  );
+  if (metrics._fromHistory) return cost;
+  const sid = sessionId || (
+    window.sessionModule && window.sessionModule.getCurrentSessionId()
+  );
+  if (!sid || cost === null) return cost;
+  const runId = typeof metrics._costRecordId === 'string'
+    ? metrics._costRecordId.trim()
+    : '';
+  if ((metrics._costRecorded || metrics._costRecordPending) && !runId) return cost;
+  // Recorded is only set once the write actually runs; pending covers the
+  // window while the write waits on the cross-tab lock, so a replay in that
+  // window cannot double-add and a tab closed mid-queue never claims recorded.
+  metrics._costRecordPending = true;
+  const writeCost = () => {
+    if (runId) {
+      try {
+        const runCosts = JSON.parse(localStorage.getItem(_COST_RUNS_KEY) || '{}');
+        const sessionRuns = runCosts[sid] && typeof runCosts[sid] === 'object'
+          ? runCosts[sid]
+          : {};
+        // Assigning by detached-run identity is replay-idempotent even when a
+        // refresh produces a fresh metrics object. The Web Lock around this
+        // read/modify/write also keeps distinct runs from two tabs from
+        // overwriting one another's stale snapshot.
+        sessionRuns[runId] = cost;
+        const entries = Object.entries(sessionRuns);
+        if (entries.length > _MAX_COST_RUNS_PER_SESSION) {
+          const overflow = entries.slice(0, entries.length - _MAX_COST_RUNS_PER_SESSION);
+          const costs = JSON.parse(localStorage.getItem(_COST_KEY) || '{}');
+          costs[sid] = (costs[sid] || 0) + overflow.reduce(
+            (total, entry) => total + (Number(entry[1]) || 0),
+            0,
+          );
+          overflow.forEach(([oldRunId]) => delete sessionRuns[oldRunId]);
+          localStorage.setItem(_COST_KEY, JSON.stringify(costs));
+        }
+        runCosts[sid] = sessionRuns;
+        localStorage.setItem(_COST_RUNS_KEY, JSON.stringify(runCosts));
+      } catch (_e) { /* ignore */ }
+    } else {
+      try {
+        const costs = JSON.parse(localStorage.getItem(_COST_KEY) || '{}');
+        costs[sid] = (costs[sid] || 0) + cost;
+        localStorage.setItem(_COST_KEY, JSON.stringify(costs));
+      } catch (_e) { /* ignore */ }
+    }
+    metrics._costRecorded = true;
+    metrics._costRecordPending = false;
+    const currentSid = window.sessionModule && window.sessionModule.getCurrentSessionId();
+    if (currentSid === sid) updateSessionCostUI();
+  };
+
+  let writeStarted = false;
+  const guardedWrite = () => {
+    writeStarted = true;
+    writeCost();
+  };
+  try {
+    if (
+      typeof navigator !== 'undefined'
+      && navigator.locks
+      && typeof navigator.locks.request === 'function'
+    ) {
+      const pendingWrite = navigator.locks.request(_COST_LEDGER_LOCK, guardedWrite);
+      if (pendingWrite && typeof pendingWrite.catch === 'function') {
+        pendingWrite.catch(() => {
+          if (!writeStarted) guardedWrite();
+        });
+      }
+    } else {
+      guardedWrite();
+    }
+  } catch (_e) {
+    if (!writeStarted) guardedWrite();
+  }
+  return cost;
+}
+
 /** Create a timestamp span for role labels.
  * Pass an ISO string / Date / epoch-ms to render the message's own time
  * (used when replaying history). Falls back to "now" when no value is given. */
@@ -1201,7 +1373,7 @@ document.addEventListener('click', function(e) {
       } catch {}
     });
   } else if (kind === 'document') {
-    import('./document.js?v=20260722emailfastindex1').then(mod => {
+    import('./document.js?v=20260815approvalsave1').then(mod => {
       const open = mod.loadDocument
         || mod.openDocument
         || (mod.default && (mod.default.loadDocument || mod.default.openDocument));
@@ -1223,7 +1395,7 @@ document.addEventListener('click', function(e) {
       if (open) open(id);
     }).catch(() => {});
   } else if (kind === 'email') {
-    import('./emailLibrary.js?v=20260722emailfastindex1').then(mod => {
+    import('./emailLibrary.js?v=20260815approvalsave1').then(mod => {
       const open = mod.openEmailLibrary || (mod.default && mod.default.openEmailLibrary);
       if (open) open({ uid: id });
     }).catch(() => {});
@@ -1382,7 +1554,7 @@ export function buildImageBubble(imageUrl, prompt, model, size, quality, imageId
     try {
       const [galleryMod, editorMod] = await Promise.all([
         import('./gallery.js'),
-        import('./galleryEditor.js'),
+        loadPanel('editor'),
       ]);
       // Ensure the Gallery modal is open so the editor has a container
       // to render into; switch its tabs to the Edit tab.
@@ -1874,23 +2046,19 @@ export function displayMetrics(messageElement, metrics) {
   const isReal = metrics.usage_source === 'real';
   const ctxPct = metrics.context_percent;
   const model = metrics.model || 'Unknown';
-  const cost = _billableCost(model, inputTokens, outputTokens);
+  const cost = _metricsBillableCost(
+    metrics,
+    model,
+    inputTokens,
+    outputTokens,
+  );
 
   // Nothing useful to show — bail out (only if ALL metrics are missing)
   if (!responseTime && !inputTokens && !outputTokens && tps == null && !ctxPct) return;
 
-  // Accumulate session cost (only on fresh metrics, not history reload)
-  if (!metrics._fromHistory) {
-    const _sid = window.sessionModule && window.sessionModule.getCurrentSessionId();
-    if (_sid && cost !== null) {
-      try {
-        const _costs = JSON.parse(localStorage.getItem(_COST_KEY) || '{}');
-        _costs[_sid] = (_costs[_sid] || 0) + cost;
-        localStorage.setItem(_COST_KEY, JSON.stringify(_costs));
-      } catch (_e) { /* ignore */ }
-      updateSessionCostUI();
-    }
-  }
+  // Rendering can occur when metrics arrive and again after [DONE]. The
+  // ledger mutation is idempotent for that shared payload.
+  recordSessionMetricsCost(metrics);
 
   // Keep token counts in the Message Stats popup; the footer should stay slim.
   const costStr0 = cost !== null ? `$${cost < 0.01 ? cost.toFixed(4) : cost.toFixed(3)}` : null;
@@ -2180,6 +2348,7 @@ export function renderAskUserCard(payload, options) {
   card.setAttribute('role', 'group');
   card.tabIndex = -1;
   const multi = !!aq.multi;
+  const isToolApproval = aq.kind === 'tool_approval' && !!aq.approval_id;
   const emojiText = (value) => svgifyEmoji(uiModule.esc(String(value)));
 
   const head = document.createElement('div');
@@ -2204,6 +2373,27 @@ export function renderAskUserCard(payload, options) {
   card.appendChild(question);
   card.setAttribute('aria-labelledby', question.id);
 
+  if (isToolApproval && aq.action) {
+    const action = document.createElement('div');
+    action.className = 'ask-user-option-desc';
+    const effects = Array.isArray(aq.action.effects)
+      ? aq.action.effects.join(', ')
+      : '';
+    action.textContent = [
+      aq.action.tool || 'tool',
+      aq.action.content || '',
+      effects ? `Effects: ${effects}` : '',
+      aq.action.workspace ? `Workspace: ${aq.action.workspace}` : '',
+      aq.action.document_id ? `Document: ${aq.action.document_id}` : '',
+      aq.action.document_version != null
+        ? `Document version: ${aq.action.document_version}`
+        : '',
+      aq.action.digest ? `Approval fingerprint: ${aq.action.digest}` : '',
+    ].filter(Boolean).join('\n');
+    action.style.whiteSpace = 'pre-wrap';
+    card.appendChild(action);
+  }
+
   const list = document.createElement('div');
   list.className = 'ask-user-options';
   card.appendChild(list);
@@ -2241,7 +2431,23 @@ export function renderAskUserCard(payload, options) {
     }
     if (!multi) {
       row.type = 'button';
-      row.addEventListener('click', () => send(label));
+      row.addEventListener('click', () => {
+        if (isToolApproval) {
+          card.remove();
+          document.dispatchEvent(new CustomEvent('odysseus:tool-approval', {
+            detail: {
+              approval_id: aq.approval_id,
+              decision: String((opt && opt.value) || '').toLowerCase(),
+              label,
+              document_id: aq.action && aq.action.document_id
+                ? String(aq.action.document_id)
+                : '',
+            },
+          }));
+        } else {
+          send(label);
+        }
+      });
     }
     list.appendChild(row);
   });
@@ -2277,7 +2483,7 @@ export function renderAskUserCard(payload, options) {
   });
   other.appendChild(otherInput);
   other.appendChild(otherSend);
-  card.appendChild(other);
+  if (!isToolApproval) card.appendChild(other);
 
   chatBox.appendChild(card);
   if (renderOptions.scroll !== false) {
@@ -2307,9 +2513,19 @@ export function addMessage(role, content, modelName, metadata) {
     const textRaw = Array.isArray(content) ? markdownModule.renderContent(content) : content;
 
     // --- Agent multi-bubble reconstruction from saved metadata ---
-    if (role === 'assistant' && metadata && metadata.tool_events && metadata.tool_events.length > 0) {
+    if (
+      role === 'assistant'
+      && metadata
+      && (
+        (Array.isArray(metadata.tool_events) && metadata.tool_events.length > 0)
+        || (Array.isArray(metadata.round_texts) && metadata.round_texts.length > 1)
+      )
+    ) {
       const roundTexts = metadata.round_texts || [];
-      const toolEvents = metadata.tool_events;
+      const roundModels = metadata.round_models || [];
+      const roundEndpointIds = metadata.round_endpoint_ids || [];
+      const roundEndpointLabels = metadata.round_endpoint_labels || [];
+      const toolEvents = metadata.tool_events || [];
       let pendingAskUser = null;
       let lastWrap = null;
       let firstMsgAi = null;
@@ -2317,16 +2533,20 @@ export function addMessage(role, content, modelName, metadata) {
 
       const toolsByRound = {};
       for (const ev of toolEvents) {
-        const r = ev.round || 1;
+        const r = ev.round ?? 1;
         if (!toolsByRound[r]) toolsByRound[r] = [];
         toolsByRound[r].push(ev);
       }
 
-      const maxRound = Math.max(...Object.keys(toolsByRound).map(Number), roundTexts.length);
+      const toolRounds = Object.keys(toolsByRound).map(Number);
+      const maxRound = Math.max(toolRounds.length ? Math.max(...toolRounds) : 0, roundTexts.length);
 
-      for (let r = 0; r < maxRound; r++) {
-        const roundNum = r + 1;
-        const txt = resolveDocumentPlaceholderLinks((roundTexts[r] || '').trim(), metadata);
+      const firstRound = (toolsByRound[0] || []).length ? 0 : 1;
+      for (let roundNum = firstRound; roundNum <= maxRound; roundNum++) {
+        const r = roundNum - 1;
+        const txt = r >= 0
+          ? resolveDocumentPlaceholderLinks((roundTexts[r] || '').trim(), metadata)
+          : '';
 
         if (txt) {
           const wrap = document.createElement('div');
@@ -2334,10 +2554,31 @@ export function addMessage(role, content, modelName, metadata) {
           const roleEl = document.createElement('div');
           roleEl.className = 'role';
           const pair = replyModelPair(modelName, metadata);
-          const contModel = pair.actualModel || pair.requestedModel;
-          roleEl.textContent = modelRouteLabel(pair.requestedModel, contModel);
-          if (pair.requestedModel && contModel && !sameModelName(pair.requestedModel, contModel)) {
-            roleEl.title = pair.requestedModel + ' -> ' + contModel;
+          const contModel = roundModels[r] || pair.actualModel || pair.requestedModel;
+          const contEndpointId = r < roundEndpointIds.length
+            ? roundEndpointIds[r]
+            : pair.actualEndpointId;
+          const contEndpointLabel = r < roundEndpointLabels.length
+            ? roundEndpointLabels[r]
+            : pair.actualEndpointLabel;
+          roleEl.textContent = modelRouteLabel(
+            pair.requestedModel,
+            contModel,
+            pair.requestedEndpointLabel,
+            contEndpointLabel,
+            pair.requestedEndpointId,
+            contEndpointId,
+          );
+          if (
+            pair.requestedModel
+            && contModel
+            && (
+              !sameModelName(pair.requestedModel, contModel)
+              || (pair.requestedEndpointId && contEndpointId && pair.requestedEndpointId !== contEndpointId)
+            )
+          ) {
+            roleEl.title = pair.requestedModel + ' -> ' + contModel
+              + ' (' + pair.requestedEndpointLabel + ' -> ' + contEndpointLabel + ')';
           }
           applyModelColor(roleEl, contModel);
           if (r === 0) roleEl.appendChild(roleTimestamp(metadata?.timestamp));
@@ -2492,7 +2733,14 @@ export function addMessage(role, content, modelName, metadata) {
     const isCompacted = metadata?.compacted;
     const replyModels = replyModelPair(modelName, metadata);
     const resolvedModel = replyModels.actualModel || replyModels.requestedModel;
-    var _roleText = role === 'user' ? 'You' : (isSlash || isCompacted) ? 'Odysseus' : modelRouteLabel(replyModels.requestedModel, resolvedModel);
+    var _roleText = role === 'user' ? 'You' : (isSlash || isCompacted) ? 'Odysseus' : modelRouteLabel(
+      replyModels.requestedModel,
+      resolvedModel,
+      replyModels.requestedEndpointLabel,
+      replyModels.actualEndpointLabel,
+      replyModels.requestedEndpointId,
+      replyModels.actualEndpointId,
+    );
     if (role === 'assistant' && (metadata?.research || metadata?.research_clarification)) {
       _roleText += ' (Research)';
     }
@@ -2503,8 +2751,14 @@ export function addMessage(role, content, modelName, metadata) {
     }
     r.textContent = _roleText;
     if (role !== 'user') {
-      if (!isSlash && !isCompacted && replyModels.requestedModel && resolvedModel && !sameModelName(replyModels.requestedModel, resolvedModel)) {
-        r.title = replyModels.requestedModel + ' -> ' + resolvedModel;
+      const endpointChanged = Boolean(
+        replyModels.requestedEndpointId
+        && replyModels.actualEndpointId
+        && replyModels.requestedEndpointId !== replyModels.actualEndpointId
+      );
+      if (!isSlash && !isCompacted && replyModels.requestedModel && resolvedModel && (!sameModelName(replyModels.requestedModel, resolvedModel) || endpointChanged)) {
+        r.title = replyModels.requestedModel + ' -> ' + resolvedModel
+          + ' (' + replyModels.requestedEndpointLabel + ' -> ' + replyModels.actualEndpointLabel + ')';
       }
       if (!isSlash && !isCompacted) applyModelColor(r, resolvedModel);
       r.appendChild(roleTimestamp(metadata?.timestamp));
@@ -2788,6 +3042,7 @@ const chatRenderer = {
   getSessionCost,
   resetSessionCost,
   updateSessionCostUI,
+  recordSessionMetricsCost,
   roleTimestamp,
   stripToolBlocks,
   copyMessageText,

static/js/chatStream.js

@@ -7,7 +7,7 @@ import Storage from './storage.js';
 import themeModule from './theme.js';
 import markdownModule from './markdown.js';
 import sessionModule from './sessions.js';
-import documentModule from './document.js?v=20260722emailfastindex1';
+import documentModule from './document.js?v=20260815approvalsave1';
 
 /**
  * Handle a ui_control SSE event — AI-driven UI manipulation.
@@ -156,7 +156,7 @@ export function handleUIControl(uiData) {
           if (fn) fn();
         }).catch(function(){});
       } else if (panel === 'email') {
-        import('./emailLibrary.js?v=20260722emailfastindex1').then(function(mod) {
+        import('./emailLibrary.js?v=20260815approvalsave1').then(function(mod) {
           var fn = mod.openEmailLibrary || (mod.default && mod.default.openEmailLibrary);
           if (fn) fn();
         }).catch(function(){});
@@ -205,7 +205,7 @@ export function handleUIControl(uiData) {
       } catch (e) {
         console.warn('open_email_reply existing draft update failed:', e);
       }
-      import('./emailInbox.js?v=20260722emailfastindex1').then(function(mod) {
+      import('./emailInbox.js?v=20260815approvalsave1').then(function(mod) {
         var fn = mod.openReplyDraft || (mod.default && mod.default.openReplyDraft);
         if (fn) fn(uiData.uid, uiData.folder || 'INBOX', uiData.mode || 'reply', uiData.body || '');
       }).catch(function(e) {

static/js/chatStreamErrors.js

@@ -0,0 +1,23 @@
+/** Build a terminal stream error while preserving provider-supplied text. */
+export function createTerminalStreamError(payload = {}) {
+  const rawError = payload.error;
+  const message = (
+    payload.text
+    || (typeof rawError === 'string' ? rawError : rawError?.message)
+    || `Error ${payload.status || 'unknown'}`
+  );
+  const error = new Error(message);
+  error.name = 'TerminalStreamError';
+  error.terminalStreamError = true;
+  error.status = payload.status;
+  return error;
+}
+
+/** Only connection-class stream failures are safe to resubmit automatically. */
+export function isRecoverableStreamError(error) {
+  if (!error || error.terminalStreamError || error.name === 'TerminalStreamError') return false;
+  if (error.name === 'TypeError') return true;
+  const message = (error.message || '').toLowerCase();
+  if (/\btool\b|unsupported|json|parse|\b4\d\d\b|\b5\d\d\b/.test(message)) return false;
+  return /network|fetch|connection|reset|closed|aborted|stream|tim(?:e|ed)\s?out|econn|eof/.test(message);
+}

static/js/document.js

@@ -3934,7 +3934,7 @@ import { bindMenuDismiss, dismissOrRemove } from './escMenuStack.js';
             leadingIcon: 'check',
             action: 'View Message',
             onAction: () => {
-              import('./emailLibrary.js?v=20260722emailfastindex1').then(mod => {
+              import('./emailLibrary.js?v=20260815approvalsave1').then(mod => {
                 const open = mod.openEmailLibrary || (mod.default && mod.default.openEmailLibrary);
                 if (open) open({
                   account_id: data.account_id || activeAccountId || null,
@@ -9401,9 +9401,9 @@ import { bindMenuDismiss, dismissOrRemove } from './escMenuStack.js';
 
   /** Save manual edits */
   export async function saveDocument({ silent = false, forceVersion = false } = {}) {
-    if (!activeDocId) return;
+    if (!activeDocId) return false;
     const textarea = document.getElementById('doc-editor-textarea');
-    if (!textarea) return;
+    if (!textarea) return false;
     const savingDocId = activeDocId;
     saveCurrentToMap();
     const localDoc = docs.get(savingDocId);
@@ -9422,7 +9422,7 @@ import { bindMenuDismiss, dismissOrRemove } from './escMenuStack.js';
       });
       if (res.status === 404) {
         if (silent && localDoc?.language === 'email') {
-          return;
+          return false;
         }
         // Streaming/empty email drafts can leave a local tab pointing at a temp
         // or already-deleted document. Do not keep surfacing autosave errors for
@@ -9434,7 +9434,7 @@ import { bindMenuDismiss, dismissOrRemove } from './escMenuStack.js';
         }
         _syncDocIndicator();
         if (!silent && uiModule) uiModule.showError('Document no longer exists');
-        return;
+        return false;
       }
       if (!res.ok) throw new Error(`Document save failed: HTTP ${res.status}`);
       const doc = await res.json();
@@ -9447,6 +9447,7 @@ import { bindMenuDismiss, dismissOrRemove } from './escMenuStack.js';
       }
       _syncDocIndicator();
       if (!silent && uiModule) uiModule.showToast(forceVersion ? 'New version saved' : 'Document saved');
+      return true;
     } catch (e) {
       console.error('Failed to save document:', e);
       const now = Date.now();
@@ -9454,6 +9455,7 @@ import { bindMenuDismiss, dismissOrRemove } from './escMenuStack.js';
         uiModule.showError(silent ? 'Autosave failed' : 'Failed to save document');
         _lastAutoSaveErrorAt = now;
       }
+      return false;
     }
   }
 
@@ -9736,6 +9738,11 @@ import { bindMenuDismiss, dismissOrRemove } from './escMenuStack.js';
     const container = document.createElement('div');
     container.style.cssText = 'padding:20px;font-family:sans-serif;font-size:12px;color:#000;background:#fff;line-height:1.6;';
     container.innerHTML = html;
+    // This container is detached, so the document-scoped flush mdToHtml
+    // schedules never sees it. Typeset the deferred math before html2pdf
+    // rasterises, or the PDF gets raw formula source. renderMath() returns
+    // immediately, without loading KaTeX, when there is nothing pending.
+    await markdownModule.renderMath(container);
     const baseName = _getExportBaseName();
     window.html2pdf().set({
       margin: 10,

static/js/emailInbox.js

@@ -5,7 +5,7 @@
 
 import spinnerModule from './spinner.js';
 import sessionModule from './sessions.js';
-import { initEmailLibrary, openEmailLibrary, closeEmailLibrary, isOpen as isLibOpen, prewarmEmailLibrary, prewarmUnreadEmails } from './emailLibrary.js?v=20260722emailfastindex1';
+import { initEmailLibrary, openEmailLibrary, closeEmailLibrary, isOpen as isLibOpen, prewarmEmailLibrary, prewarmUnreadEmails } from './emailLibrary.js?v=20260815approvalsave1';
 import * as Modals from './modalManager.js';
 import { applyEdgeDock } from './modalSnap.js';
 import { buildReplyAllCc, extractEmail } from './emailLibrary/replyRecipients.js';

static/js/emailLibrary.js

@@ -5,7 +5,7 @@
 
 import spinnerModule from './spinner.js';
 import { styledConfirm, showToast, emptyStateIcon } from './ui.js';
-import { folderDisplayName, sortedFolders } from './emailInbox.js?v=20260722emailfastindex1';
+import { folderDisplayName, sortedFolders } from './emailInbox.js?v=20260815approvalsave1';
 import settingsModule from './settings.js';
 import * as Modals from './modalManager.js';
 import { topPortalZ } from './toolWindowZOrder.js';
@@ -23,6 +23,7 @@ import {
   _tryFoldHintSig, _foldSignature, _SIG_ICON, _QUOTE_ICON,
 } from './emailLibrary/signatureFold.js';
 import { state } from './emailLibrary/state.js';
+import { getSettings } from './appConfig.js';
 import { collapseSidebarToRail } from './modalSnap.js';
 import { emailApiUrl } from './emailShared.js';
 import { bindMenuDismiss, dismissOrRemove } from './escMenuStack.js';
@@ -993,8 +994,7 @@ function _syncEmailReminderBellVisibility(enabled) {
 
 async function _loadEmailReminderBellVisibility() {
   try {
-    const res = await fetch('/api/auth/settings', { credentials: 'same-origin' });
-    const settings = await res.json();
+    const settings = await getSettings();
     _syncEmailReminderBellVisibility(settings.reminder_channel === 'email');
   } catch (_) {
     _syncEmailReminderBellVisibility(false);
@@ -6680,7 +6680,7 @@ function _wireAttachmentHandlers(reader, folder) {
               ownerModal.classList.add('hidden');
             }
           }
-          const docMod = await import('./document.js?v=20260722emailfastindex1');
+          const docMod = await import('./document.js?v=20260815approvalsave1');
           const load = (docMod && docMod.loadDocument) || (docMod && docMod.default && docMod.default.loadDocument);
           if (typeof load === 'function') {
             await load(json.doc_id);

static/js/gallery.js

@@ -3,7 +3,7 @@
  */
 
 import uiModule from './ui.js';
-import { openEditor, closeEditor, isEditorOpen } from './galleryEditor.js?v=20260708match1';
+import { loadPanel } from './panels.js';
 import spinnerModule from './spinner.js';
 import { makeWindowDraggable } from './windowDrag.js';
 import { bindMenuDismiss, dismissOrRemove } from './escMenuStack.js';
@@ -15,6 +15,54 @@ const API_BASE = window.location.origin;
 let _open = false;
 let _galleryResizeHandler = null;
 
+// ── Image editor, loaded on first use ──
+// galleryEditor.js plus everything under js/editor/ is 54 modules / 576 KB.
+// It used to be a static import here, so every page load paid for it even
+// though most sessions never touch the Edit tab. The wrappers below keep the
+// three call shapes the rest of this file already uses.
+//
+// closeEditor() and isEditorOpen() stay synchronous on purpose: if the module
+// was never loaded there is no edit session to close, and none can be open.
+let _editorMod = null;
+let _editorLoading = false;
+
+async function _loadEditor() {
+  _editorLoading = true;
+  try {
+    _editorMod = await loadPanel('editor');
+    return _editorMod;
+  } finally {
+    _editorLoading = false;
+  }
+}
+
+async function openEditor(...args) {
+  let mod = _editorMod;
+  if (!mod) {
+    try {
+      mod = await _loadEditor();
+    } catch (e) {
+      // Previously unreachable — a static import either loaded or the whole
+      // page failed. Now it can fail on its own (offline before the panel was
+      // ever cached), so say so instead of doing nothing.
+      console.error('[gallery] image editor failed to load', e);
+      uiModule?.showError?.('Failed to load the image editor');
+      return;
+    }
+  }
+  return mod.openEditor(...args);
+}
+
+function closeEditor(...args) {
+  return _editorMod ? _editorMod.closeEditor(...args) : undefined;
+}
+
+// True while the module is still in flight as well — the gallery-close paths
+// use this to refuse to tear the container down under an edit that is opening.
+function isEditorOpen() {
+  return _editorLoading || (_editorMod ? _editorMod.isEditorOpen() : false);
+}
+
 // Auto-refresh gallery when new image is generated
 window.addEventListener('gallery-refresh', (e) => {
   if (e?.detail?.source === 'chat-upload' && _sort !== 'recent') {

static/js/keyboard-shortcuts.js

@@ -3,6 +3,7 @@
 // ============================================
 
 import { IS_MAC, isAltGrEvent } from './platform.js';
+import { getSettings } from './appConfig.js';
 
 const _defaultKeybinds = {
   search: 'ctrl+k', toggle_sidebar: 'ctrl+alt+b', new_session: 'ctrl+alt+n',
@@ -56,8 +57,7 @@ export function initKeyboardShortcuts(modules) {
   window._odysseusKeybinds = { ..._defaultKeybinds };
 
   // Load saved keybinds
-  fetch('/api/auth/settings', { credentials: 'same-origin' })
-    .then(r => r.json())
+  getSettings()
     .then(s => { if (s.keybinds) window._odysseusKeybinds = { ..._defaultKeybinds, ...s.keybinds }; })
     .catch(() => {});
 

static/js/markdown.js

@@ -10,6 +10,127 @@ import { replaceEmojiShortcodes, hasEmojiShortcode } from './emojiShortcodes.js'
 
 var escapeHtml = uiModule.esc;
 
+// Mermaid and KaTeX are vendored under /static/lib and fetched on first use.
+// Loading them from <head> cost every session ~985 KB on the wire even though
+// most chats never contain a diagram or a formula. Both loaders memoise the
+// *promise* rather than the resolved library, so concurrent callers share one
+// fetch and a double trigger cannot start two loads. A failed load clears the
+// memo so the next diagram/formula retries instead of being poisoned forever.
+const MERMAID_SRC = '/static/lib/mermaid.min.js';
+const KATEX_SRC = '/static/lib/katex/katex.min.js';
+const KATEX_CSS = '/static/lib/katex/katex.min.css';
+// Marks math emitted before KaTeX finished loading; renderMath() swaps these
+// for typeset output. The source stays as readable text inside the span, so a
+// load that never completes degrades to plain text rather than to nothing.
+const MATH_PENDING_CLASS = 'ody-math-pending';
+
+// KaTeX has no entity syntax: it reads a bare "&" as an alignment marker and
+// errors out on anything that is not a valid column break, so "a &lt; b" comes
+// back as a red .katex-error instead of a formula. mdToHtml escapes the whole
+// string before the math pass, which leaves two spellings of the same
+// character at the delimiters — a typed "<" arrives as "&lt;", while a typed
+// "&lt;" arrives as "&amp;lt;" — and both have to reach KaTeX as "<".
+//
+// One alternation, longest form first, so nothing this writes is scanned
+// again. Chained .replace() calls cannot do it: unescaping "&amp;" first lets
+// the next pass eat the "&lt;" it just produced (the double-unescape CodeQL
+// flags), and unescaping it last leaves the entity spelling intact and breaks
+// the render. The code-block pass upstream keeps its chained order on purpose
+// — Markdown does not decode entities inside code, so "&lt;" there is meant to
+// stay visible.
+const MATH_SOURCE_ENTITY_RE = /&amp;(?:lt|gt|amp|quot|#39);|&lt;|&gt;|&amp;/g;
+const MATH_SOURCE_ENTITIES = {
+  '&amp;lt;': '<',
+  '&amp;gt;': '>',
+  '&amp;amp;': '&',
+  '&amp;quot;': '"',
+  '&amp;#39;': "'",
+  '&lt;': '<',
+  '&gt;': '>',
+  '&amp;': '&',
+};
+
+function decodeMathSource(text) {
+  return String(text).replace(MATH_SOURCE_ENTITY_RE, (entity) => MATH_SOURCE_ENTITIES[entity]);
+}
+
+let _mermaidPromise = null;
+let _katexPromise = null;
+let _mathFlushScheduled = false;
+
+function _loadScript(src) {
+  return new Promise((resolve, reject) => {
+    const script = document.createElement('script');
+    script.src = src;
+    script.addEventListener('load', () => resolve(), { once: true });
+    script.addEventListener('error', () => reject(new Error('Failed to load ' + src)), { once: true });
+    document.head.appendChild(script);
+  });
+}
+
+function _loadStylesheet(href) {
+  // Resolves either way: without the stylesheet KaTeX still produces correct
+  // markup, just unstyled, which beats failing the whole math render.
+  return new Promise((resolve) => {
+    const link = document.createElement('link');
+    link.rel = 'stylesheet';
+    link.href = href;
+    link.addEventListener('load', () => resolve(), { once: true });
+    link.addEventListener('error', () => resolve(), { once: true });
+    document.head.appendChild(link);
+  });
+}
+
+/**
+ * Load Mermaid on first use and initialize it once.
+ */
+export function ensureMermaid() {
+  return (_mermaidPromise ??= _loadScript(MERMAID_SRC)
+    .then(() => {
+      if (!window.mermaid) throw new Error('mermaid global missing after load');
+      window.mermaid.initialize({ startOnLoad: false, theme: 'dark', securityLevel: 'loose' });
+      return window.mermaid;
+    })
+    .catch((err) => {
+      _mermaidPromise = null;
+      throw err;
+    }));
+}
+
+/**
+ * Load KaTeX (script + stylesheet) on first use.
+ */
+export function ensureKatex() {
+  return (_katexPromise ??= Promise.all([_loadScript(KATEX_SRC), _loadStylesheet(KATEX_CSS)])
+    .then(() => {
+      if (!window.katex) throw new Error('katex global missing after load');
+      return window.katex;
+    })
+    .catch((err) => {
+      _katexPromise = null;
+      throw err;
+    }));
+}
+
+// mdToHtml() is synchronous and its callers insert the returned string into the
+// DOM themselves, so the placeholders are usually not attached yet when this
+// fires. Loading first and scanning afterwards covers that gap: by the time
+// KaTeX is in, the caller's innerHTML assignment has long since happened.
+//
+// setTimeout, not requestAnimationFrame: this has nothing to do with paint, and
+// rAF is throttled to a stop in a background tab (and never fires at all in a
+// headless browser), which would leave math untypeset until the tab is focused.
+function _scheduleMathFlush() {
+  if (_mathFlushScheduled) return;
+  _mathFlushScheduled = true;
+  setTimeout(() => {
+    _mathFlushScheduled = false;
+    ensureKatex()
+      .then(() => renderMath(document))
+      .catch((e) => console.warn('KaTeX load error:', e));
+  }, 0);
+}
+
 function safeLinkUrl(rawUrl) {
   const url = String(rawUrl || '').trim();
   if (url.startsWith('#')) {
@@ -631,49 +752,45 @@ export function mdToHtml(src, opts) {
 
   // KaTeX math rendering (after code blocks are extracted, so math in code is safe)
   const mathBlocks = [];
-  if (window.katex) {
-    // Display math: \[ ... \]  — GPT-style delimiter (gpt-5.x, Claude, etc.).
-    // Handle before $$/$ so all common delimiters render.
-    s = s.replace(/\\\[([\s\S]*?)\\\]/g, (match, math) => {
-      try {
-        const raw = math.replace(/&amp;/g, '&').replace(/&lt;/g, '<').replace(/&gt;/g, '>');
-        const placeholder = `___MATH_BLOCK_${mathBlocks.length}___`;
-        mathBlocks.push(katex.renderToString(raw.trim(), { displayMode: true, throwOnError: false }));
-        return placeholder;
-      } catch (e) { return match; }
-    });
-    // Inline math: \( ... \)  — GPT-style inline delimiter. Single-line only
-    // ([^\n]) so a stray escaped paren in prose can't swallow across lines.
-    s = s.replace(/\\\(([^\n]*?)\\\)/g, (match, math) => {
-      try {
-        const raw = math.replace(/&amp;/g, '&').replace(/&lt;/g, '<').replace(/&gt;/g, '>');
-        const placeholder = `___MATH_BLOCK_${mathBlocks.length}___`;
-        mathBlocks.push(katex.renderToString(raw.trim(), { displayMode: false, throwOnError: false }));
-        return placeholder;
-      } catch (e) { return match; }
-    });
-    // Display math: $$...$$
-    s = s.replace(/\$\$([\s\S]*?)\$\$/g, (match, math) => {
-      try {
-        const raw = math.replace(/&amp;/g, '&').replace(/&lt;/g, '<').replace(/&gt;/g, '>');
-        const placeholder = `___MATH_BLOCK_${mathBlocks.length}___`;
-        mathBlocks.push(katex.renderToString(raw.trim(), { displayMode: true, throwOnError: false }));
-        return placeholder;
-      } catch (e) { return match; }
-    });
-    // Inline math: $...$ — single line only, and Pandoc-style delimiter rules so
-    // currency doesn't render as math ("$5 to $10"): the opening $ must be
-    // immediately followed by a non-space, the closing $ must be immediately
-    // preceded by a non-space and not followed by a digit.
-    s = s.replace(/(?<![\$\d])\$(?!\$)(?=\S)([^\$\n]+?)(?<=\S)\$(?!\$|\d)/g, (match, math) => {
-      try {
-        const raw = math.replace(/&amp;/g, '&').replace(/&lt;/g, '<').replace(/&gt;/g, '>');
-        const placeholder = `___MATH_BLOCK_${mathBlocks.length}___`;
-        mathBlocks.push(katex.renderToString(raw.trim(), { displayMode: false, throwOnError: false }));
-        return placeholder;
-      } catch (e) { return match; }
-    });
-  }
+  let sawPendingMath = false;
+
+  // Typeset straight away when KaTeX is already in, otherwise bank the source in
+  // an inert placeholder for renderMath() to swap once the library lands.
+  const pushMath = (math, displayMode) => {
+    const raw = decodeMathSource(math).trim();
+    const placeholder = `___MATH_BLOCK_${mathBlocks.length}___`;
+    if (window.katex) {
+      mathBlocks.push(katex.renderToString(raw, { displayMode, throwOnError: false }));
+    } else {
+      sawPendingMath = true;
+      mathBlocks.push(`<span class="${MATH_PENDING_CLASS}" data-display="${displayMode}">${escapeHtml(raw)}</span>`);
+    }
+    return placeholder;
+  };
+
+  // Display math: \[ ... \]  — GPT-style delimiter (gpt-5.x, Claude, etc.).
+  // Handle before $$/$ so all common delimiters render.
+  s = s.replace(/\\\[([\s\S]*?)\\\]/g, (match, math) => {
+    try { return pushMath(math, true); } catch (e) { return match; }
+  });
+  // Inline math: \( ... \)  — GPT-style inline delimiter. Single-line only
+  // ([^\n]) so a stray escaped paren in prose can't swallow across lines.
+  s = s.replace(/\\\(([^\n]*?)\\\)/g, (match, math) => {
+    try { return pushMath(math, false); } catch (e) { return match; }
+  });
+  // Display math: $$...$$
+  s = s.replace(/\$\$([\s\S]*?)\$\$/g, (match, math) => {
+    try { return pushMath(math, true); } catch (e) { return match; }
+  });
+  // Inline math: $...$ — single line only, and Pandoc-style delimiter rules so
+  // currency doesn't render as math ("$5 to $10"): the opening $ must be
+  // immediately followed by a non-space, the closing $ must be immediately
+  // preceded by a non-space and not followed by a digit.
+  s = s.replace(/(?<![\$\d])\$(?!\$)(?=\S)([^\$\n]+?)(?<=\S)\$(?!\$|\d)/g, (match, math) => {
+    try { return pushMath(math, false); } catch (e) { return match; }
+  });
+
+  if (sawPendingMath) _scheduleMathFlush();
 
   // Handle pipe tables
   s = s.replace(/(?:^|\n)([^\n]*\|[^\n]*\|[^\n]*)(?:\n([^\n]*\|[^\n]*\|[^\n]*))*/g, (table) => {
@@ -826,19 +943,47 @@ export function renderContent(content) {
 }
 
 /**
- * Initialize any unprocessed Mermaid diagrams in a container (or whole document)
+ * Initialize any unprocessed Mermaid diagrams in a container (or whole document).
+ * Returns a promise so callers can await the (lazy) library load if they need to.
  */
 export function renderMermaid(container) {
-  if (!window.mermaid) return;
-  initMermaid();
   const target = container || document;
-  const pending = target.querySelectorAll('pre.mermaid:not([data-processed])');
-  if (pending.length === 0) return;
-  try {
-    window.mermaid.run({ nodes: pending });
-  } catch (e) {
-    console.warn('Mermaid render error:', e);
-  }
+  if (!target || typeof target.querySelectorAll !== 'function') return Promise.resolve();
+  // Cheap pre-check: no fence on the page means Mermaid is never fetched.
+  if (target.querySelectorAll('pre.mermaid:not([data-processed])').length === 0) return Promise.resolve();
+  return ensureMermaid()
+    .then((mermaid) => {
+      // Re-query after the load: during streaming the renderer replaces the
+      // message body repeatedly, so the nodes seen before the fetch are stale.
+      const nodes = [...target.querySelectorAll('pre.mermaid:not([data-processed])')]
+        .filter((node) => node.isConnected);
+      if (nodes.length === 0) return;
+      return mermaid.run({ nodes });
+    })
+    .catch((e) => { console.warn('Mermaid render error:', e); });
+}
+
+/**
+ * Typeset any math that mdToHtml() had to defer because KaTeX was not loaded
+ * yet. Once KaTeX is in, mdToHtml() renders inline and this finds nothing.
+ */
+export function renderMath(container) {
+  const target = container || document;
+  if (!target || typeof target.querySelectorAll !== 'function') return Promise.resolve();
+  if (target.querySelectorAll('.' + MATH_PENDING_CLASS).length === 0) return Promise.resolve();
+  return ensureKatex()
+    .then((katex) => {
+      target.querySelectorAll('.' + MATH_PENDING_CLASS).forEach((el) => {
+        const displayMode = el.getAttribute('data-display') === 'true';
+        try {
+          el.outerHTML = katex.renderToString(el.textContent || '', { displayMode, throwOnError: false });
+        } catch (e) {
+          // Leave the source visible — readable, just not typeset.
+          el.classList.remove(MATH_PENDING_CLASS);
+        }
+      });
+    })
+    .catch((e) => { console.warn('KaTeX render error:', e); });
 }
 
 const markdownModule = {
@@ -853,20 +998,14 @@ const markdownModule = {
   extractThinkingBlocks,
   normalizeThinkingMarkup,
   startsWithReasoningPrefix,
-  renderMermaid
+  renderMermaid,
+  renderMath,
+  ensureMermaid,
+  ensureKatex
 };
 
 export default markdownModule;
 
-// Mermaid is loaded async so it cannot delay the app shell.
-function initMermaid() {
-  if (!window.mermaid || window.__odysseusMermaidReady) return;
-  window.mermaid.initialize({ startOnLoad: false, theme: 'dark', securityLevel: 'loose' });
-  window.__odysseusMermaidReady = true;
-}
-window.odysseusInitMermaid = initMermaid;
-initMermaid();
-
 // Persist which thinking sections were expanded across page refreshes.
 // IDs are render-generated (Date.now-based) so we key by a stable hash of
 // the inner text content instead — same content reproduces the same hash on

static/js/panels.js

@@ -0,0 +1,53 @@
+/**
+ * Panel loader registry — imports the modules behind a feature panel the
+ * first time that panel is actually used.
+ *
+ * The panels already populate themselves on open (each fetches its own data).
+ * They were just not *loaded* on demand: every one of them sat on the critical
+ * path of every page load, opened or not.
+ *
+ * A panel only belongs here once it has been checked for import-time side
+ * effects that something outside the panel depends on at startup. Entries get
+ * added one panel at a time, not in bulk.
+ *
+ * Note the service worker still precaches these modules (PANEL_PRECACHE in
+ * sw.js) — they are off the critical path, not off the offline manifest.
+ */
+
+const LOADERS = {
+  editor: () => import('./galleryEditor.js'),
+};
+
+/**
+ * Build a memoising loader over a name -> import-thunk map. Exported so the
+ * behaviour can be tested without pulling a real panel's module graph in.
+ */
+export function createPanelLoader(loaders) {
+  const cache = new Map();
+  return function load(name) {
+    const cached = cache.get(name);
+    if (cached) return cached;
+    const loader = loaders[name];
+    if (!loader) throw new Error(`loadPanel: unknown panel "${name}"`);
+    // A failed load (offline, 404, syntax error) is not memoised — caching the
+    // rejection would leave the panel broken for the rest of the session even
+    // after the network came back.
+    const pending = Promise.resolve().then(loader).catch((err) => {
+      cache.delete(name);
+      throw err;
+    });
+    cache.set(name, pending);
+    return pending;
+  };
+}
+
+/**
+ * Load a panel's module, once. Returns the same promise on every call for the
+ * same panel; throws for a name that is not registered.
+ */
+export const loadPanel = createPanelLoader(LOADERS);
+
+/** Registered panel names — the registry is the list, not a second copy of it. */
+export function panelNames() {
+  return Object.keys(LOADERS);
+}

static/js/search.js

@@ -4,20 +4,21 @@
  * Search settings management — reads active provider from admin settings.
  */
 
-let API_BASE = '';
+import { getSettings, invalidateSettings } from './appConfig.js';
+
 let _provider = 'searxng';
 let _loaded = false;
 
-export function init(apiBase) {
-  API_BASE = apiBase;
+// No API base parameter any more: the settings request lives in appConfig.js and
+// resolves against the document origin, which is exactly what API_BASE held.
+export function init() {
   // Fetch provider on init so it's ready when chat needs it
   _fetchProvider();
 }
 
 async function _fetchProvider() {
   try {
-    const res = await fetch((API_BASE || '') + '/api/auth/settings', { credentials: 'same-origin' });
-    const s = await res.json();
+    const s = await getSettings();
     _provider = s.search_provider || 'searxng';
     _loaded = true;
   } catch (e) { /* keep default */ }
@@ -39,6 +40,9 @@ export function getProviderLabel() {
 
 /** Re-fetch after admin saves new settings */
 export function refresh() {
+  // Drop the shared snapshot first: the point of this call is to observe the
+  // settings that were just written, so it must not be served from cache.
+  invalidateSettings();
   _fetchProvider();
 }
 

static/js/sessions.js

@@ -3,7 +3,7 @@
 
 import Storage from './storage.js';
 import uiModule, { autoResize, styledPrompt } from './ui.js';
-import chatRenderer from './chatRenderer.js?v=20260722ctxheader1';
+import chatRenderer from './chatRenderer.js?v=20260815toolapproval4';
 import { providerLogo } from './providers.js';
 import { initModelPicker, updateModelPicker } from './modelPicker.js?v=20260722ctxheader1';
 import themeModule from './theme.js';