refactor(model-routing): centralize explicit foreground fallback policy (#6020)

* refactor(model-routing): centralize explicit foreground fallback policy

Make foreground fallback an explicit per-user, availability-only policy shared by streaming Chat, non-stream Chat, and Agent runs.

Preserve strict defaults, owner/model and credential boundaries, pinned Agent routes, and truthful per-round provenance/accounting. Carry provider-reported model identifiers through native streaming adapters, non-stream responses, and caches, and keep legacy default_model_fallbacks as tombstoned raw storage that generic settings APIs and agent tools cannot expose or mutate.

* fix(agent-loop): restore rebase-dropped qwen routing, workspace prompt, and temperature clamp

* fix(model-routing): thread selected endpoint identity, fix cost classification and fallback eligibility

* fix(chat): restore stream helpers and harden run stop lifecycle

* fix(model-routing): let numeric provider codes win over symbolic rate-limit statuses

* fix(agent-loop): apply qwen temperature and notes-tool clamps per fallback candidate

* fix(chat): honor queued stop across resend and reload canonical terminal on EOF

* fix(chat): track stop queue and cleanup ownership by per-send generation

* fix(agent-loop): preserve requested temperature for non-qwen fallback candidates

* fix(chat): reserve send ownership before any await and scope stop to the current send

* fix(chat): clear the previous run identity at send reservation

---------

Co-authored-by: RaresKeY <158580472+RaresKeY@users.noreply.github.com>
Co-authored-by: StressTestor <212606152+StressTestor@users.noreply.github.com>

tests/test_context_compactor.py

@@ -63,6 +63,23 @@ class TestSelfSummaryPrompt:
 
 
 class TestTrimForContext:
+    def test_system_truncation_preserves_internal_route_metadata(self):
+        messages = [
+            {
+                "role": "system",
+                "content": "persona\n\n" + ("agent prompt " * 2000),
+                "_agent_injected": "merged_prompt",
+                "_agent_base_message": {"role": "system", "content": "persona"},
+            },
+            {"role": "user", "content": "latest"},
+        ]
+
+        trimmed = trim_for_context(messages, context_length=1024, reserve_tokens=256)
+
+        system = next(message for message in trimmed if message.get("role") == "system")
+        assert system["_agent_injected"] == "merged_prompt"
+        assert system["_agent_base_message"] == {"role": "system", "content": "persona"}
+
     def test_keeps_current_large_user_message_by_truncating(self):
         huge = "A" * 20000
         messages = [
@@ -194,6 +211,50 @@ class TestMaybeCompactFourthMessage:
         assert len(result) == 3 and result[2] is True
 
 
+@pytest.mark.asyncio
+async def test_deferred_compaction_persists_only_after_route_commit(monkeypatch):
+    updates = []
+    state = {}
+    messages = [
+        {"role": "system", "content": "system " * 100},
+        {"role": "user", "content": "one"},
+        {"role": "assistant", "content": "two"},
+        {"role": "user", "content": "three"},
+        {"role": "assistant", "content": "four"},
+        {"role": "user", "content": "five"},
+    ]
+
+    monkeypatch.setattr(cc, "get_context_length", lambda *args: 100)
+    monkeypatch.setattr(cc, "resolve_endpoint", lambda *args, **kwargs: (None, None, None))
+
+    async def fake_summary(*args, **kwargs):
+        return "route-specific summary"
+
+    monkeypatch.setattr(cc, "llm_call_async", fake_summary)
+    monkeypatch.setattr(
+        cc,
+        "_update_session_history",
+        lambda *args, **kwargs: updates.append((args, kwargs)),
+    )
+
+    _compacted, _context, was_compacted = await cc.maybe_compact(
+        object(),
+        "https://candidate.example/v1",
+        "candidate-model",
+        messages,
+        persist=False,
+        compaction_state=state,
+    )
+
+    assert was_compacted is True
+    assert updates == []
+    assert state["summary"] == "route-specific summary"
+    assert cc.apply_compaction_state(object(), state) is True
+    assert len(updates) == 1
+    assert cc.apply_compaction_state(object(), state) is False
+    assert len(updates) == 1
+
+
 class TestResearchPrimerPreserved:
     """A research-spinoff primer (metadata research_spinoff_from) must never be
     trimmed away — it is the Discuss chat's sole knowledge base (drift fix)."""