Phase 1 follow-up: update the active instruction layer (CLAUDE.md, project
structure rule, DECISIONS.md redirect, agent personalities, skill docs) to
reference governance/{decisions,questions,rejected}/<domain>.md instead of the
retired flat decisions/*.md layout.
Path references only — command-surface references (tooling/db/decision*,
decisions-sync, Makefile targets, clerk) are repointed to the pql CLI in the
Phase 4 consumer cutover. Historical archives (docs/sprints, docs/discussions,
docs/workshops) keep their point-in-time decisions/ paths; the separate
whatsinagame/ template distribution is untouched. Agent-memory is gitignored
and out of scope.
The agent/skill repath was applied by tooling/pql-migrate/repath_references.py
(ordered, meaning-preserving replacements; bare-dir rule uses a negative
lookbehind so it can't corrupt a freshly-created governance/decisions/ path),
committed for provenance. CLAUDE.md, project-structure.md, and DECISIONS.md
were hand-edited (structural tree/table changes).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Phase 1 of the pql migration. Moves the flat decisions/*.md layout into
governance/{decisions,questions,rejected}/<domain>.md — the tree pql's
`decisions sync` parses natively (record type from subdir, domain from
filename stem). Proven against pql 1.6.2: sync reports 357 records
(237 D / 108 Q / 12 R), 1057 refs, broken: 0; validate ok.
- 6 D-domain files -> governance/decisions/ (git renames)
- 5 questions-<domain>.md -> governance/questions/<domain>.md (prefix dropped)
- rejected.md split by domain -> governance/rejected/{architecture(R-001..010),
economics(R-011),perception(R-012)}.md
- decisions/README.md + questions.md index folded into governance/README.md;
pql's `decisions sync` now auto-maintains the record index appended below
the hand-written domain guidance (no more manual ID-list table upkeep).
- .pql/config.yaml: canonical vault config (tracked, not ignored).
Link rewrites are token-preserving: only the relative `foo.md` path portion
changes (e.g. `rejected.md#r-011` -> `../rejected/economics.md#r-011`); every
`[D-NNN]` bracket text and `#anchor` stays byte-identical, so pql's reference
extraction is unaffected. The one-shot transform is committed at
tooling/pql-migrate/restructure_decisions.py for provenance.
Codebase path references to decisions/ (CLAUDE.md, rules, skills, docs) are
updated in a follow-up commit.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Phase 0 of the pql migration: the decision markdown had two genuine
canonical-ID integrity bugs that pql decisions sync silently LWW-collapses.
- D-035 was defined twice: a content.md tag-taxonomy record and a
perception.md shadowcasting/LOS record, both dated the same day. The
tag-taxonomy record keeps D-035; the shadowcasting record is renumbered
to D-238 (freshly claimed) with a Renumbered note preserving the history.
Inbound refs repointed in architecture.md (x4), questions-architecture.md,
and the README perception index row.
- R-011 (single currency for Phase 2) was defined three times: once
canonically in rejected.md and twice identically in economics.md. The two
economics.md copies are removed; rejected.md absorbs the Proposal/Raised-by
lines and gains a dedupe note. The economics.md dissent link now points at
rejected.md and the footer count drops the rejected-alternative clause.
Verified: tooling/db/decision check-dupes reports 0 duplicates;
pql decisions validate returns ok.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The blanket `.pql/` ignore would have hidden `.pql/changelog/` — pql's durable,
git-versioned planning state — making the whole migration's output invisible to
git. Narrow to ignore only the rebuildable DBs (index/pql .db + WAL/SHM) and
track config.yaml + changelog/.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Deep analysis of migrating decisions + ticketing from the SQLite CLI to pql:
- pql-migration.md: phased branch-only plan (decisions→DQR tree, ticket changelog
seed with T-N≡#N, #N→T-N find-replace, big-bang consumer cutover, docs/workshops/
wiki fold-in, SQLite retirement) + verification gate + benefits.
- pql-requirements.md: 9 surfaced gaps for the pql team (critical: seeded ticket
ids + counter-advance; high: dup-id detection, core.hooksPath awareness, changelog
seed-format docs).
Key finding: milestones are vestigial (milestone_deps empty, 42/1013 tickets
linked) → mapped to a label, dropped as an entity.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
PR #153 review (Hoshe H1): zone_type/era/floors/flavor all shared one fp_chain
and each took splitmix64(seed) of the same value, correlating the fields (era
tracking floor height across a city). Derive a distinct sub-chain per field so
they draw independent entropy. Block_tags aren't serialized, so no fixture
change; behavior stays deterministic.
Also (review): n=1 corridor test (empty, no panic), Frontier-density waterfront
test, and a comment on the reserved `_morphology` param (Tyre).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Note the #957 implementation on D-234: typed AccessPoint/CorridorSpine/
ChunkLayout, morphology-gated arterials + per-block lattice, BSP footprint
subdivision with D-233 coverage, and the waterfront rule. Flag the remaining
cross-layer dependency: the waterfront rule's real input (per-settlement Coastal
orientation) needs the Layer-3 placement -> Layer-4 GenerateSkeleton dispatch,
since city_context_reader still stubs founding_orientation to Cardinal.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Close the last D-234 piece — terrain water-direction extraction wired through
to founding orientation and the quarter waterfront rule:
- Layer 1: TerrainAnalysis::water_bearing — 8-octant integer bearing toward the
nearest water from the water_dist gradient (D-010, no atan2). Stored on
GeographicAttractor.water_bearing (360 = none).
- #956 founding orientation: coastal/river settlements now get a real
water-facing bearing (the anchoring attractor's), replacing the 0 stub.
- #957 waterfront rule (D-234b): the water-facing quarter edge (from the
settlement's Coastal founding orientation) drops its block setback to 0 so
buildings present flush to the quay (dock-orthogonal). Typed Edge + coastal_edge
+ per-block gating.
Golden + atlas_response fixture rebaked (additive water_bearing field only).
8 new tests. All integer-deterministic (D-010).
Pending: the waterfront rule reads context.founding_orientation, which
city_context_reader still stubs to Cardinal — real per-settlement orientation
reaches quarter generation once the Layer-3 placement -> Layer-4 GenerateSkeleton
dispatch is wired (the remaining cross-layer integration).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Complete the Layer-4 quarter geometry with the street layer (D-234), replacing
the ChunkLayout/CorridorSpine/AccessPoint String stubs with real types:
- AccessPoint{position,kind} + AccessKind (QuarterEdge/ReservationGate/
BlockJunction): the quarter's road nodes (from road_entry_directions octants +
reservation gates), the node set the D-097 audit reads.
- CorridorSpine{from,to,path}: arterial trunk edges, ±45°-snapped polylines
(D-096 cap). Topology gated by morphology (D-234a): Ribbon (fjord/canyon/
mountain-pass), HubSpoke (delta/island/enclosed water), Mesh/Prim-MST
(plains/meander/coastal). A spanning tree over the access nodes.
- ChunkLayout{spacing,offset,rotation_steps}: per-block local ±45° lattice,
spacing from density, offset/rotation modulated by the D-096 Grid/Organic mode.
Wired into generate_quarter_skeleton (access_points + corridors + per-block
chunk_layout). All integer-deterministic (D-010). 7 new tests.
Remaining D-234 piece: the per-edge waterfront pier/quay rule needs Layer-1
terrain water-adjacency threaded to the skeleton (cross-layer plumbing).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Build the per-quarter building geometry + tags (absorbs #976's footprint
subdivision into #957 — one coherent walkable-quarter deliverable):
- subdivide_block_footprints: deterministic BSP of each 128x128-tile block into
variable axis-aligned building plots (D-229 fast-path). Lot size + setback
scale with D-220 density (Frontier -> few big lots/wide gaps; Compressed ->
many small lots/shared walls); the D-233 BulkClass roofed-coverage fraction
decides building-vs-interstitial. morphology is wired in for the D-234
waterfront/street layer (needs Layer-1 terrain water-adjacency threaded up).
- assign_all_block_tags: subdivides + tags every non-reserved block's footprints
via the #957 derivation helpers, populating QuarterWorldState.block_tags in the
GenerateSkeleton plan-phase (D-230). Reserved blocks get no standard fill.
- Remove the write-only BlockSkeleton.era String stub - construction era now
lives per-footprint on BuildingPropertyTag.era (typed ConstructionEra).
All integer-deterministic (D-010). 5 new tests. Doors stay Vec::new() (#979).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Record the (ZoningType × economic_role × setting) → ZoneTypeId selection table
for the planetary cascade: base candidate slices per ZoningType refined by
economic_role, a setting tweaker (Maritime/Agricultural/Wilderness — planetary
variants only), and the gap-fill rules (entry-class U-curve at the D-217 Broken
band <2300bps; era per founding_age+prosperity+seed; FloorExtent from D-220
density midpoints; doors deferred to #979). Station-only zone ids are excluded —
they belong to a separate cascade (Q-109).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The pure per-footprint tag derivations for Layer-4 quarter generation (D-229),
ahead of the footprint-subdivision + assembly step:
- zone_type_for: the authored (ZoningType × economic_role × setting) → ZoneTypeId
selection (D-229 amendment) — planetary variants only, setting as a tweaker,
station-only ids never selected (Q-109); deterministic seed-pick, no empty slice.
- building_entry_class: zone × layout_mode × prosperity, with the D-217 Broken-band
U-curve degrade (Commercial→BreachOnly) and the Grid/Organic credential fork.
- construction_era: founding_age + prosperity + seed (D-229); old+Broken → Derelict.
- floor_extent: D-220 density-class floor ranges from density_pct + seed-jitter.
- initial_condition: D-217 frozen-amber condition from prosperity + era cause.
All integer-deterministic (D-010). 9 unit tests incl. the no-station-ids invariant.
Footprint subdivision + assign_building_tags assembly + block_tags wiring follow
(needs the footprint-packing model decided).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Surfaced during #957 zone-selection authoring: a body's setting/mode should
dispatch the cascade entry to planetary / station / mod-DLC-forked /
save-only(player_base) sources, distinct from setting's in-path tweaker role.
Captured as an open question so #957 stays scoped to the planetary path; this is
why #957's zone table excludes station-only zone ids.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
PR #152 review (Hoshe): the faction→TerritorialStatus mapping test covered 7 of
the 8 D-237 values; add the missing "mixed" assertion (grouped with "disputed",
behavior already correct).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- D-212: add the AutonomistHeld variant (Compact of Westphalia) and record the
dominant_faction → TerritorialStatus mapping used in place of the
numeric faction-influence thresholds (which aren't in the data). Grounded in
wiki/factions/. IndigenousHeld/Derelict noted as deferred.
- D-214: note the political_archetype derivation + per-settlement storage.
- D-215: note the ArrangementPattern enum + derivation land here; block-adjacency
enforcement stays with the Quarter-skeleton generator (#957).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
DrainageBasin gained a territorial_status field, so the Layer-1 golden and the
atlas_response msgpack fixture pick it up. Both changes are additive/
representational — basin/attractor/river selection is unchanged.
- cascade_golden.rs: pass None dominant_faction; golden rebaked (only
"territorial_status": "FrontierUnclaimed" added per basin).
- gen_fixtures.rs: set the new field in the atlas_response generator; fixture
regenerated.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Derive and store each placed settlement's spatial character on top of the
#955 placement (D-212/213/214/215). Derivation is pure + integer-deterministic
(D-010); the cascade stays DB-free (D-225) — the enqueuer pre-resolves the
body's system faction onto the work item, like #955's settlements.
- TerritorialStatus (D-212): mapped from the authored dominant_faction
(territorial_status_from_faction). Adds an AutonomistHeld variant for the
Compact of Westphalia (self-governing bloc that rejects Assembly authority —
neither Commission, Corp, Contested, nor truly Frontier). Grounded in
wiki/factions/. Stored per-province on DrainageBasin.territorial_status
(uniform per body for now; forward-compatible for per-province faction data).
- PoliticalArchetype (D-214): political_archetype(status, role), status takes
precedence over economic_role.
- ArrangementPattern (D-215): new 5-variant enum + arrangement_pattern(); the
block-adjacency *enforcement* stays deferred to the Quarter-skeleton gen (#957)
— this only derives + stores which pattern applies.
- FoundingOrientation (D-213): existing fn extended with a seed-derived Free
bearing so pioneer/open-terrain grids vary per seed.
Wiring: dominant_faction threaded through CityContextReader
(read_body_dominant_faction) → atlas proxy (cache-miss read) → AnalyzeBody work
item → run_cascade → run_layer3/match_cities. Enrichment stored on CityPlacement
(political_archetype, arrangement_pattern, founding_orientation).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
PR #151 review (Hoshe/Tyre): four doc comments in gen_queue.rs still said
"district" next to the renamed quarter_id field / generate_quarter_skeleton.
Pure comment update, no code change.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
D-222 renamed the 512m generation cell from District to Quarter (District
is now a new 2048m tier above it). Align the generation skeleton code to
the canonical vocabulary. Pure naming — no behavior change; all 1296 lib
tests + integration tests pass unchanged.
Renamed (spatial-cell identifiers):
- DistrictSkeleton → QuarterSkeleton, DistrictWorldState → QuarterWorldState
- DistrictId → QuarterId, DistrictContext/DistrictBoundaries → Quarter*
- field district_id → quarter_id, district_type → quarter_type
- BodyWorldState.districts map → .quarters
- generate_skeleton → generate_quarter_skeleton
Deliberately left as-is (these name functional ZONING, not the spatial
tier — orthogonal to D-222): DistrictType, DistrictLayoutMode, the
district_mix module (DistrictMix/compute_district_mix), and the
GenWorkItem::GenerateSkeleton / GenCompletion::SkeletonGenerated variants.
Also aligned the perception "sim tile" → "subtile" vocabulary (D-222:
Subtile = 0.5m) in decisions/perception.md and the generation-cascade code
comments. Historical D-066/D-094/D-201/D-220 decision bodies keep their
existing D-222 amendment notes (not rewritten in place); the public
max_offset_sim_tiles fn name is unchanged.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Per-machine tool dirs (.clide/settings.yaml, .pql/pql.db) that appeared as
untracked noise; not shared, should never be committed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The doctest imported `server::bps::log10_floor`, but the library crate is
`settled_reach_server` — so `cargo test --doc` failed to compile the
example (unresolved module `server`). Broken since the helper landed in
#145; it only surfaces under `--doc`, not `--lib`, so the pre-push gate
(which runs `--lib`) never caught it.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address the doc/test-correctness items from PR #149 review:
- cascade.rs: Layer 3 is RNG-free (pure fn of attractors+cities); it does
not consume the carried SeedChain. Corrected the "first RNG-using layer"
claims on the module doc and the Settlement variant (Hoshe H3).
- cascade.rs: layers_are_ordered now also asserts Topography < Settlement,
pinning the invariant the up_to >= Settlement guards rely on (Tyre T2).
- features.rs: strength is an integer 0-100 (quantized at extraction), not
f32, and consumers that rank by it use the integer value (Hoshe H2).
- layer_proxy.rs: AtlasLayerRequest.up_to is not yet honored — run_work_item
runs through Settlement unconditionally; per-request depth deferred to
#1021 (Tyre T1).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
read_body_settlements propagated a parse error for any non-NULL but
unrecognized settlement_class, which the atlas proxy swallowed and then
enqueued the body with zero cities — one malformed row silently dropped
placement for the entire body. Treat an unknown variant like NULL: fall
back to PopulationBudget with a warning. parse_settlement_class stays
strict for the D-199 read_set path, which must abort on bad fields.
Found in PR #149 review (Hoshe H1). Adds a test covering the fallback.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mark Layer-3 settlement placement (#955) as running in-game now that the
B/E/F plumbing (per-body settlement read → work item → cache-miss proxy)
is wired.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wire the existing D-211 attractor-matching engine into the live
generation cascade so settlements are placed in-game, not just in tests.
- CityContextReader::read_body_settlements reads a body's settlements
from atlas_city_names (ordered by id for determinism). NULL
settlement_class defaults to PopulationBudget, not NameLocked: the
class is NULL until placement runs, and NameLocked would force every
settlement Tier-A in match_cities and collapse population tiering
(D-211). NULL economic_role falls back to residential.
- The AnalyzeBody work item carries the body's Vec<CityRecord>, and
run_work_item now runs up_to Settlement (was Topography). A body with
no settlements yields empty placements at negligible cost.
- The atlas layer proxy reads settlements on a cache miss and pins them
onto the work item, keeping the Rayon task DB-free (D-225). A read
failure is non-fatal: log and place no cities (Layer 1 still runs).
Threaded through a new CityContextReaderResource Bevy resource opened
in main.rs, mirroring BodySourceResolverResource.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
all_fixtures_deserialize panicked with "unknown fixture naming
convention: atlas_response_ready" — the test reads every .msgpack in
client/tests/fixtures/msgpack but had no branch for the atlas_response
prefix (the AtlasLayerResponse fixtures gen_fixtures emits, #969). A
pre-existing gap (present on main; the suite isn't gated by the pre-push
hook, which runs cargo-deny/ruff/JSON/stamp, not cargo test).
Add an atlas_response branch deserializing AtlasLayerResponse, and
regenerate atlas_response_ready.msgpack so its bytes match the now-integer
GeographicAttractor types (#955).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wire the existing attractor-matching engine (#919/#925) into the
generation cascade as Layer 3, and make the whole placement-scoring path
integer-deterministic.
Layer 3 (D-211):
- CascadeLayer::Settlement + Layer3Output (placements) on the snapshot;
BodyWorldState gains a `placements` field (the D-203 hot cache).
- run_layer3 runs the five-phase match_cities against Layer-1 attractors
via the authored D-195 compatibility matrix; pure function of
(attractors, cities) — no RNG. cities are passed in by the caller so the
cascade stays DB-free and testable. A `// cache seam` marks where a
persistent cache wraps it later (#1021).
- gen_queue passes &[] for now (Topography needs no cities); the runtime
settlement read (gen_queue/layer_proxy) is the #955 follow-on.
Integer determinism (D-010 / D-227 — D-195 amended):
- Wiring match_cities into the deterministic cascade made its f32 scoring
a live cross-platform divergence risk (a near-tie comparison or the
Hungarian's f32 reductions can round differently per platform → a
different world from the same seed). Converted the entire path to
integers: CompatibilityMatrix is a 0-100 affinity table; attractor
strength is 0-100 and terrain cost is a percent (100 = baseline),
quantized once at the Layer-1 feature boundary; cell_score, the
Hungarian, and CityPlacement.score are i64. No f32 in any placement or
ranking decision.
- Layer-1 golden fixture rebaked: confirmed selection/positions are
unchanged (same 256 attractors, 93 river cells) — only the strength/cost
representation changed.
Tests: lib green (1292); new settlement_layer_places_cities_deterministically
covers placement + determinism + propagation into BodyWorldState.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Review fixes (Hoshe/Tyre, both APPROVE):
- populate_trait_templates / populate_atlas_body_trait_bias: the table
DELETEs ran unconditionally (safe only via transaction rollback on
dry-run, and divergent from every other populate_* function). Restructured
so validation + guardrails always run (dry-run now actually surfaces the
would-bake counts and catches errors) but mutations happen only under
`if not dry_run:`. Verified: --dry-run reports 28 templates, writes nothing.
- atlas_body_trait_bias: reject `pin` entries that carry a
weight_multiplier_bps (pin is mandatory, no multiplier) — closes a silent-
accept gap before #1017 authors ~30-40 real pins.
Deferred (noted on tickets): visual_bundle fallback-map completeness
(Phase-5/Araminta), pin-count-vs-K bake check (#1017 acceptance),
geographic_sector pool-narrowing semantics (#977).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The D-232 core hand-curated catalog (28 templates) the architecture-flavor
draw reads from, baked into trait_templates (#993):
- Cross-corridor pool (10): economic-function templates gated by BulkClass
(extraction/industrial/cold-chain/precision/information/civic) + universal
baselines + the foreign-import swerve. Guarantees the CI floor.
- Per-corridor baseline (8): each corridor's default cohesive look — core
cosmopolitan, north anglo-frontier, south lusophone, west germanic +
compact-cooperative, east dense-utilitarian, deep-frontier surname +
hardscrabble.
- Heritage sub-pools (10): deep-history callbacks tied to the D-237 heritage
taxonomy (scottish highland, west-african compound, iberian hacienda,
atlantic creole, nordic timber, central-european blok, east-asian temple,
vietnamese water village, afrikaans kraal, arab oasis).
Each template = holistic bundle: two-tier eligibility (hard gates
bulk_class/ubiquity/prosperity_bps + soft weight mods), zone_affinity over
the real DistrictType enum, allow/block tags from a shared ObjectTag
palette, and a D-235 visual_bundle with generic fallback parents. All
numerics integer basis-points (D-010).
Adds the D-232 CI guardrails to the baker (V-TT-01: >=5 templates eligible
per BulkClass; V-TT-02: no template >60% of pool weight) — both pass with
margin (23-24 eligible/class, 5% max share). Catalog can grow via the
bounded Gemma pass (#992); hero pins are #1017.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Review fixes for the #1016 cultural pass (Hoshe/Paula/Miri):
- Kruger 60 (GJ 860A) xhosa -> cosmopolitan (Miri): GTTR describes an
Afrikaner/Xhosa/Zulu founding "without mapping cleanly onto any of
them" — the textbook genuinely-fused case, which taxonomy rule 4
resolves to cosmopolitan rather than one of the three. Commented.
- Altmark (GJ 357) financial_technocratic -> cosmopolitan (Paula):
reverts to the original shipped value; the GTTR reads as commercial
non-alignment, not credential-based technocracy.
- Kampala Gate (GJ 667B) + Moyale (GJ 1292) keep swahili but add inline
comments documenting that these are East-African-INTERIOR founders
(Baganda; Kenyan/Ethiopian Borana-Oromo) for which the taxonomy has no
dedicated token — swahili is the nearest available, NULL would fall to
the wrong corridor default. Gap logged for a follow-up taxonomy value.
Mwangaza (693) swahili is literally correct (coastal) — untouched.
- heritage-taxonomy-draft.md: stale "(38)" subheadings -> "(47)" to match
the actual canonical count (Hoshe/Miri doc typo).
Retracted: Pedra Seca (GJ 421B) xhosa — reviewer thought it lacked
authored backing, but its gttr.md explicitly states "Afrikaans and Xhosa
heritage"; pin kept.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Completes the deferred cultural half of #1016 and migrates the heritage
vocabulary to the canonical 47-value taxonomy (D-237).
Migration (mechanical):
- _CULTURAL_HERITAGE replaced with the canonical 47-value set
(docs/.../heritage-taxonomy-draft.md). Renamed the 4 shipped hero
values to canonical: afrikaans_cape->afrikaans, french_provencal->
french, italian_northern->italian, norse_compact->nordic.
Content pass (authored from GTTR/name heritage triage, 6-sector fan-out):
- +139 new cultural_specialization pins and 1 change (Altmark
cosmopolitan->financial_technocratic), taking coverage 32 -> 171
across 54 distinct values. Each pin is grounded in an explicit GTTR
founding-community statement or system/feature name etymology, applied
only where founding heritage DIVERGES from the corridor baseline
(D-167/D-232); corridor-typical systems left NULL.
Distribution is now balanced and diverse — portuguese 15, german 12,
vietnamese 10, afrikaans 10 (no longer dominant post-#1019), then a long
tail filling the previously-missing slots (welsh, herero, czech, akan,
hungarian, konkan, afro_brazilian, cape_verdean, shona, arab/persian/
turkic, etc.).
Judgment calls (flagged for review):
- Kruger 60 + Pedra Seca -> xhosa (mixed SA founding; adds diversity).
Pedra Seca has a Portuguese name but Afrikaans/Xhosa GTTR founding —
name/heritage mismatch noted for a future content fix.
- Kampala Gate / Moyale / Mwangaza -> swahili as the nearest token for
East-African-interior heritage; taxonomy may want a dedicated value.
- Kept hero pins Kensho=scholarly, Keid=scholarly, Nyrheim=nordic.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
GJ 680 has aperture_count=2 / gate_connections=2, linking two distinct
systems (Athabasca GJ 1289 and Enugu GJ 75) — but the prose, inherited
verbatim from the original Carnarvon page, described "one gate in, one
gate out, both the same gate" and "one aperture connecting to Enugu and
Athabasca" (self-contradictory). Reword the gttr and index to two gates
while keeping the system's low-through-traffic spur character (the
spur_end topology label and "nobody routes through here" identity are
unchanged). Resolves the deferred PR #146 review finding.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Review fixes for the #1019 heritage rebalance:
- Marib's farm world Ghouta -> Ghawr (Miri): "Ghouta" carries strong
modern associations (2013 Damascus attack); Ghawr (the Jordan Rift
Valley, an irrigated lowland farming region) keeps the water-fed-
drylands sense without the resonance.
- Add dominant_faction = "independent" to GJ 914A / 508A / 1245B (Hoshe)
— all five rebalanced systems are self-governing independent
communities; uniform with siblings GJ 722 / 680 and clears V-FAC-01.
- catalog.md: fill GJ 914A -> Marib and GJ 722 -> Sığınak (Hoshe) — the
two anchor systems still showed em-dash placeholders.
- Marib gttr: name the second aperture neighbour Sığınak instead of the
bare "GJ 722" (Miri).
Pre-existing Golestan aperture/spur wording (reviewers split on whether
it is even wrong) is left for a separate content pass — not introduced
by this rename.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The pre-execution conversion plan and shipped-state addendum for the
South-African heritage rebalance: per-system field maps, the worked
gttr rewrites, the cross-reference cleanup list, and the rationale for
keeping load-bearing mystery systems out of scope.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The corridor over-represented southern-African heritage (afrikaans on 15
systems; arab=2, persian/turkic=0) as an artifact of the earlier
name-balancing overcorrecting. Convert five genuinely-thin afrikaans
systems — whose distinctive threads are heritage-agnostic — into
Arab/Persian/Turkic founders so the cultural cascade is built on a
balanced set. Load-bearing mystery systems (Eerste Wacht / Helderoog /
Brandpunt etc.) are left untouched.
Droëland (GJ 914A) -> Marib (arab)
Stilwater (GJ 508A) -> Sawad (arab)
Koeberg (GJ 1245B) -> Akhgar (persian)
Skuilplek (GJ 722) -> Siginak (turkic)
Carnarvon (GJ 680) -> Golestan (persian; Mostert dynasty -> Farahani)
Result: afrikaans 15->10; arab 0->2, persian 0->2, turkic 0->1.
Full narrative retouch (not just names): gttr.md + index.md prose,
body/station proper_names, cultural_specialization, gttr_hook,
atlas_city_names, and the four brand-corps named after the converted
systems (corp IDs kept stable; only display names + products renamed).
Cross-references in neighbour systems, the catalog, the aggregate
drifter guides, and the atlas proposals are updated to match.
arab/persian/turkic added to _CULTURAL_HERITAGE so the values validate;
the full canonical 47-value migration remains #1016.
Also fixes a pre-existing canon bug: Sawad (GJ 508A) no longer claims
GJ 914A is "unsettled" — it is Marib, a three-century settlement.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Reconciles the ~30 inconsistent heritage proposals from the 6-sector #1016
content pass into one canonical cultural_specialization heritage vocabulary
(47 values). Key reconciliations: norse_compact->nordic (un-fuses people from
faction), spanish_iberian+spanish_hispanic->spanish, brazilian_portuguese+
lusophone_frontier->brazilian, italian_northern->italian, afrikaans_cape->
afrikaans, french_provencal+french_canadian->french. No compound values; one
token per system (blends take dominant heritage or the cosmopolitan activity
value).
Maintainer decisions (2026-06-01): keep ALL values distinct (no lumping —
welsh/afro_brazilian/cape_malay/xhosa/luxembourgish/konkan all retained);
keep empty-but-complete slots (yoruba/hausa/hungarian/persian/turkic). Fills
the Middle-East gap the pass exposed: arab/persian/turkic added even at near-
zero current usage so the canon is complete.
Logs the southern-African overrepresentation (~25/300, afrikaans=15) as a DLC
content-rewrite rebalancing task (#1019) rather than lumping the taxonomy.
Migration note included: 4 already-shipped hero cultural values use old suffix
forms (afrikaans_cape, french_provencal, italian_northern, norse_compact) and
will be renamed in the culture-write pass when _CULTURAL_HERITAGE is replaced.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The 6-sector content pass (core pilot + deep_frontier/north/east/south/west,
~260 systems read from both index.md and gttr.md) authored economic_
specialization + dominant_faction for the systems whose lore makes those
identities distinct. Conservative pin bar: corridor-typical systems left NULL
for the heuristic/derivation fallback.
Totals now: economic 170 (27 hero + 143), faction 76 (27 hero + 49),
cultural still 27 (DEFERRED — the culture pass needs a canonical heritage
vocabulary first; the agents surfaced ~30 real heritages from the names plus a
genuine Middle-East/Arabic gap, to be reconciled into one taxonomy before
writing cultural_specialization).
regen-db green, stamp fresh, all CI guardrails pass. Soft warnings are the
expected D-177 MonopolySource review list, the known Arbour estate/breadbasket
question, and sparse-corp_presence W-FAC-03 advisories — none blocking.
Notable pins: Haodu financial_hub/syndic_dominant, Matamba legal_archive,
Inhambane/Sonnwend/etc breadbasket, the terroir cluster, the west_reach
Compact zone, Aurelius governance_center/veil_institute. FLAGs left NULL
pending human calls (ISPI=veil_institute?, Gate-Corp-as-faction).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
generate_infobox() now renders economic_specialization (Specialization) and
cultural_specialization (Cultural Register) as read-only infobox rows when
authored (NULL = omitted, falls to heuristic). dominant_faction already
showed via Governance. Makes the authored generator-input values visible on
the wiki face so the #1016 content pass output renders. Verified: Ran shows
'breadbasket'/'agrarian'; unauthored systems omit the rows.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Records that the two are a deliberate parallel representation, not a
duplicate: economic_base_primary is uncomputable GTTR prose flavor rendered
to the wiki infobox; economic_specialization is the higher-fidelity enum the
D-233 generator consumes. Notes they must stay mutually consistent (#1016
authors the enum from the same lore; W-SES-03 flags divergence), that
unification was considered and rejected (each carries info the other can't),
and clarifies the DB-canonical / index.md-projection data flow (only the
prose sections are authored in-place). Surfaced when the content-pass review
questioned the redundancy.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
make regen-db with the D-237 authored layer live: 27-row
specialization_vocabulary, economic_specialization + cultural_specialization
on 27 hero systems, dominant_faction on the authored set. Meta stamp fresh
(check-systems-db OK). Unauthored systems remain NULL pending the #1014
fallback and #1016 content pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The #1015 V-SES-03 edit replaced the vocabulary validation loop body and
dropped the trailing vocab_rows.append(), so specialization_vocabulary was
inserted with 0 rows even though per-system UPSERTs (which read the dict
directly via valid_spec_ids) still succeeded. regen-db reported
'vocab 0 | economic 27 | ...'. Restored the append at the end of the loop.
Verified: regen-db now writes 27 vocab rows + 27/27 economic/cultural;
spot checks financial_hub=NonPhysical/Specialist, Ran=breadbasket.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Adds the V-SES / V-FAC guardrail suite to import_system_specialization():
Always-on hard errors (abort the import transaction): V-SES-01 economic
value in vocabulary; V-SES-03 production_ubiquity_override >= commodity
catalog concentration (HUB values with empty override exempt — they
intentionally project lower); V-SES-04 cultural value in the combined
activity+heritage vocabulary; V-SES-05 faction in 8-value vocab; V-SES-06
vocab commodity FK; plus unknown system_id.
Completeness gates — V-SES-02 (every inhabited system resolves a non-null
economic value) and V-FAC-01 (every inhabited named system has authored
faction) — are HARD only under --strict-specialization (default off). Their
preconditions are the #1014 heuristic fallback (blocked by #982) and the
#1016 content pass; until those land they print as warnings so regen-db
stays green.
Soft warnings (always print): W-SES-01 MonopolySource D-177 review list,
W-SES-02 single-value concentration, W-SES-08 estate_farming + large pop,
W-FAC-01 compact + TRACTUS currency, W-FAC-02 compact + MonopolySource
(terroir/marble exempt), W-FAC-03 syndic_dominant w/o corp HQ, W-FAC-04
compact_sympathetic + MARK currency. Coverage report always prints
economic/cultural/faction counts + BulkClass/ProductionUbiquity dists.
Cultural vocabulary is _CULTURAL_ACTIVITY | _CULTURAL_HERITAGE module
constants — #1016 adds heritage values there in one place.
Validated on a copy of the live DB: default exits 0 (gates as warnings, real
W-SES-08 hit on Arbour); strict exits 1 on V-SES-02 (Sol, Struve) and rolls
back cleanly; dry-run exits 0.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Wires the D-237 authored specialization layer into import_economics.py:
- import_system_specialization(): reads specialization_vocabulary.toml
(FK-validated against commodities; repopulates specialization_vocabulary)
and system_specialization.toml (UPSERTs economic_specialization +
cultural_specialization onto system_economy, dominant_faction onto
system_factions for authored systems only). Hard errors on bad commodity
FK / projected enum / unknown system_id abort the transaction; prints a
coverage report + missing-row warnings.
- Called as step 4b in main() (after commodities so the FK resolves, before
currency zones). economic/cultural columns are importer-owned and cleared
to NULL first for idempotency; dominant_faction only overwritten for
authored systems (never globally cleared — shared with other derivation).
- specialization_vocabulary added to MIGRATION_SQL (after commodities for FK)
so the migration path on existing DBs creates the table, not just fresh
systems-schema.sql builds — this is what #1011 missed. Also cleared before
commodities in the FK-safe clear block.
- Both source TOMLs added to IMPORT_ECONOMICS_SOURCES and mirrored in
check-systems-db-stamp GENERATOR_SOURCES so editing them flips the stamp.
Validated: full non-dry-run import on a copy of the live systems.db exits 0;
27 vocab rows, 27/27/27 economic/cultural/faction set, no missing rows;
spot checks Ran=breadbasket/agrarian, Vuurkloof=independent,
financial_hub=NonPhysical/Specialist. Dry-run also exits 0.
V-SES CI guardrail suite remains #1015; code note flags that V-SES-03
(equal-or-higher) must not hard-fail HUB specializations.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Per maintainer call — 3 was too tight for the parallel Phase 4 work in
flight; 5 better matches how many tickets are genuinely active at once.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Completes ticket #1010. The D-237 record and D-233 re-amendment landed
earlier; this adds the three downstream amendments:
- D-199: read-set gains economic_specialization + cultural_specialization;
dominant_faction now authored-sourced where present.
- D-214: dominant_faction is an authored 8-value vocabulary for named
systems (heuristic fallback otherwise); lattice_commission is not a
faction value (Commission is a regulator). Archetype mapping unchanged
but now reads a trustworthy faction for the systems the heuristic got
wrong (e.g. Groombridge → Syndic/Corporate, not Commission).
- D-232: template draw reads authored cultural_specialization to select/
bias the pool when founding heritage diverges from the corridor baseline
(Phase-4 correctness, not flavor); NULL = existing corridor algorithm.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The D-233/D-237 built-world fill model is implicitly a surface model
(roofed-coverage fraction, operations surface, scatter-vs-concentrate on
open ground), but bodies.settlement_pattern already carries enclosed
morphologies (underground_concentrated, cave, domed, underground_complex)
that nothing in the chain reads — not in the D-199 read-set, not in
city_context_reader, not in skeleton_gen. Engine z-level primitives exist
(D-110, D-106, UndergroundComplex) but no morphology switch connects them
at the built-world layer, so a body like Vuurkloof (GJ 35) would generate
as a surface town against its underground-concentrated lore.
Q-108 schedules the surface-vs-enclosed morphology debate before the
Phase 4 content pass authors built form for enclosed-habitat bodies.
Tracked by ticket #1018. Surfaced via Vuurkloof during the
system-economic-specialization workshop.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add the D-237 authored specialization layer (and D-233 re-amendment) under
[Unreleased].
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Running notes for a later workshop-start pass: read-only participant agents
can't self-write output files (lead-as-scribe relay needed); SI has no
round-1 task; create+assign tasks before spawning; TaskCreate is one-per-call;
decision claim is not idempotent (caused a stray duplicate ID); the optional
scout-critic idea and its guardrails; and background always-present teammates
can be slow to consume shutdown_request at wrap-up.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Three-round design workshop (Miri, Burnelli-Sheldon, Paula; documented by
Qatux, ticketed by SI) that produced D-237. Includes per-agent round
outputs, round notes, the independent scout critique, and the final
workshop-outcomes.md with the consolidated schema, 27-value economic
vocabulary, cultural + faction vocabularies, deterministic-varied fallback
spec, must-pin table, and CI guardrails.
Clerk-Skip: workshop discussion docs, no D-record surface (D-237 filed separately)
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add D-237: each inhabited system carries three authored, static (t=0),
deterministic per-system identity fields sourced from
wiki/economics/system_specialization.toml — economic_specialization
(27-value vocabulary, scale encoded in the value), cultural_specialization
(unified activity + heritage vocabulary), and dominant_faction (populate
the existing system_factions column, 8 values, lattice_commission dropped).
Authored lore wins; a deterministic-varied seed-hashed weighted draw is the
fallback for unnamed systems so they don't read identically.
Re-amend D-233 to source dominant_bulk_class x dominant_production_ubiquity
from this authored layer (heuristic becomes fallback). Groombridge encodes
as a financial_hub Specialist district with the GSH clearing house as a
D-222/D-232 landmark rather than a composite vocabulary value.
Add the specialization-layer d2 diagram (3 fields -> D-233/D-232/D-214
consumers). Produced by the system-economic-specialization workshop.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- settings.json: add Read(repo/**), Read(/tmp/**), Grep, Glob to the
allowlist. Teammates inherit settings.json (not settings.local.json),
which had no Read rule — so every teammate file read escalated to the
lead as a per-file prompt. This closes that gap.
- Add the system-economic-specialization workshop brief (lore-authored
per-system dominant-commodity field + within-system weighted variation;
Q5 covers adding a small identity field set in the same 300-system pass).
Note: agent memory-strip (6c93b698f) + this read-allow reach teammates
only after a Claude Code restart (settings/agent-def snapshot is taken at
lead startup).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Address PR #145 review. Convert prosperity from f32 to integer basis
points (matching the existing prosperity_bps, D-010 integer-only),
removing both float-determinism bugs by construction:
- New server/src/bps.rs: log10_floor (integer order-of-magnitude) +
bps_to_f32 edge helper, with boundary tests.
- city_context_reader: prosperity_baseline_bps (u32, 0-10000). role base
+ pop bonus (400*log10_floor, cap 1200) + symmetric noise via integer
modulo (fixes the always-negative bug) + clamp [1000,9500]. No floats.
- CityGenerationContext.prosperity_baseline -> prosperity_baseline_bps;
updated the two test context builders. Not serialized — no wire break.
- district_mix population_tier now uses log10_floor (same determinism
bug class as the comment claimed to avoid).
- Tests in bps + assert positive noise is achievable (the case the old
test hid).
cargo check/clippy --all-targets -D warnings clean; 1291 lib tests pass;
fmt clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Only Qatux (the librarian — cross-session recall is its job) keeps
project memory. Removed memory: project from the 20 ephemeral
implementer/reviewer/design agents that are spawned per-task and die:
they have no curation discipline and already use decisions/ + briefings/
as shared memory, so an uncurated per-agent store only invited rot.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Layer 2 clean half. GenerateSkeleton now builds a CityGenerationContext
from a body's real economic read-set instead of stub defaults.
- New CityContextReader (server/src/atlas/city_context_reader.rs): reads
the 6-field D-199 set (economic_role, prosperity_baseline, population,
dominant_faction, founding_age_years, settlement_class) via one JOIN;
prosperity_baseline derived per D-197.
- GenWorkItem::GenerateSkeleton carries the resolved context; run_work_item
builds the real context instead of DistrictWorldState::default().
- Left at defaults (deferred, documented): dominant_bulk_class /
dominant_production_ubiquity (#982, design-blocked), trait_selection
(#1005), morphology_zone + Layer-3 fields (attractor placement / D-204 /
D-214).
16 new tests; cargo check/clippy --all-targets -D warnings clean; 1279
lib tests pass; fmt clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- pr-review §7a: after a PR merges, recommend tearing down the worktree
and deleting the merged topic branch (local + origin) by default —
state the action, let the user decline; never touch protected branches;
git branch -d as the unmerged-work safety net.
- whats-next §3c: document the worktree lifecycle — created at activation,
removed by pr-review §7a at merge; prune stale worktrees before creating.
Closes the gap that left empty worktrees parked at main's HEAD.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Update D-216 amendment text to match the implemented decision:
events-first uses event_type='colonial_charter' (9 systems, 11 bodies,
5 diverging from wave) — not the stale founding/first_settlement/
colonization scaffolding vocabulary.
- Add ORDER BY sort_order ASC to the colonial_charter subquery for a
deterministic pick if a system ever has two charter rows.
- Add explicit ELSE NULL to the wave-fallback CASE so an unexpected
settlement_wave leaves a trace instead of silently nulling.
Data unchanged: 269/273 set, 4 Sol NULL; stamp fresh.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- Add D-230 skeleton: DistrictSkeleton field to DistrictWorldState;
cascade Default to DistrictSkeleton + contained enums/structs. Box the
GenCompletion::SkeletonGenerated state to avoid large_enum_variant.
- Derive PartialEq on FloorExtent/FloorHeightProfile/DistrictWorldState/
CityGenerationContext (+ minimal cascade) for downstream assert_eq tests.
- Add 4 unit tests for floor_at_voxel_z / voxel_range_for_floor (uniform,
basement, variable heights, boundary) — the Q-104 deliverable.
- Drop unused smallvec direct dep (stays transitive via bevy_ecs).
- Key districts insert by skeleton.district_id, sharpen TODO(#957).
clippy --all-targets -D warnings clean; 1263 lib tests pass; fmt clean.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Define the type-definition layer the Phase-4 fill-seam tickets depend on
(D-229/D-230/D-231/D-232/D-233), compiling with stubs/defaults; behavior
logic lands in #982-985/#998.
- New types in generator.rs: BuildingPropertyTag, FloorExtent +
FloorHeightProfile (floor_at_voxel_z/voxel_range_for_floor, resolves
Q-104), BuildingEntryClass, ConstructionEra, ZoneTypeId, MorphologyZone,
BulkClass(5), ProductionUbiquity, DoorSpec, InteriorDescriptor,
DistrictWorldState.
- Rename spatial AccessTier -> ZoneAccessTier to free the name for the new
per-building BuildingEntryClass.
- CityGenerationContext: +morphology_zone, +trait_selection,
+dominant_bulk_class, +dominant_production_ubiquity.
- BodyWorldState: +districts (DistrictWorldState w/ block_tags).
- GenCompletion::SkeletonGenerated carries body_id + DistrictWorldState;
plugin handler inserts into BodyWorldState.districts.
- Add smallvec as a direct dep (DoorSpec list stays Vec for now, TODO).
cargo check --all-targets / clippy clean; 1259 lib tests pass.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Anchors the events-first / wave-fallback COALESCE that #1000 will write into import_economics.py MIGRATION_SQL. Founding-edge values from systems-framework-final-miri.md: wave_1=600, wave_2=500, wave_3=300, wave_4=100, wave_5=40, origin=NULL (Sol excluded per D-236). historical_events is currently empty so events-first is forward-compatible scaffolding; wave-fallback is the practical default for all 273 inhabited bodies today.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Formalises what was previously implicit in D-223's naming exemption and GJ-0's wiki frontmatter into a single canonical scope rule. (a) Phase-4+ generation pipelines exclude Sol bodies (settlement_wave='origin' gate); (b) Atlas renders Sol at system level only -- no zoom; (c) deeper Sol treatment is reserved for a future DLC. Cross-refs D-145 (DLC precedent), D-191 (Atlas Phase-3), D-223.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
From the Si refinement of the Phase-4 fill-seam tickets. D-097: the mandatory-check gate keys on ComplexityTier, not WorldTier (wording fix). D-233: bulk_class is the 8->5 built-form projection (the 8 cargo-types stay on commodities); a settlement's dominant commodity is derived from its production chains, not an authored role->commodity table. D-234: streets are two-layer — MST arterials (Civ6-style, the guarantee-audit graph) + +/-45deg grid local streets (Civ4-style); Voronoi rejected (breaks +/-45 + the axis-aligned footprint fast-path). Ticket-level refinements (foundational types task, era-gating fix, catalog resequence) live in the ticket DB.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The Stage-2 workshop brief and the consolidated three-scan coverage map (per-step decided-vs-open across the 4-step built-world pipeline) that became round 2's foundation.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
atlas_gen_open opens the real atlas app via HudGroups, navigates to GJ71c's
regional screen, and enables the Layer-1 overlays — exercising the production
opener + the live request→cascade→render pipeline against a --test-mode server
(SR_LIVE=1). Capture with: tests/run-visual --screenshot atlas_gen_open.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Generation overlays map positions from Layer1Output.grid_w/grid_h onto the
displayed texture (not the markers.json grid), fixing the off-by-2 scale.
- AtlasViewer loads ImplantPending by path instead of referencing the new
class_name, so a stale global-class cache (a session that hasn't re-imported
after the class was added) can't fail to parse AtlasViewer and break the
atlas from opening.
- Rivers and mouths use the reliefmap's own water colour (sampled median ocean
blue) so river lines blend into surface water instead of reading as a
distinct-coloured line flowing onto the sea.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two Layer-1 generation fixes surfaced by the per-layer atlas viewer:
- Layer1Output now carries grid_w/grid_h (the downsampled working-grid the
positions live in). The client maps overlays from these, so the scale is
correct for any source heightmap resolution rather than assuming the texture
size — fixes overlays projecting at half scale into a corner.
- Drainage basin boundaries are traced as ordered, non-self-crossing contours
via Moore-neighbour tracing instead of an angle-from-centroid sort. The sort
produced star-shaped, self-crossing polygons for concave basins that rendered
as straight chords across the map.
Golden (cascade_layer1.json) and the cross-language atlas_response_ready
fixture regenerated. 100 atlas lib tests + the new tracer test pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Ad-hoc capture scenario that overlays the ImplantPending "generating" indicator
on the booted scene for visual sign-off — `tests/run-visual --screenshot
implant_pending`. No golden committed; this is an inspection scenario, not a
regression gate.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
visual_capture.gd runs via `godot -s`, where the `class_name` registry isn't
populated — so the bare `Protocol.decode_snapshot()` reference failed to
compile, breaking ALL --screenshot/--movie/golden captures (not just the
replay scenarios that use it). Instantiate the script and call the static
decoder on the instance (then free), matching the file's existing -s-mode
load() workaround for VisualScenarios.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
A reusable implant-layer busy indicator (Araminta's spec): an animated
bracket sweep `[ >>>>···· ]` with a caption, themed entirely via ImplantTheme.
The Atlas shows it centered over the canvas while the Layer-1 proxy is Pending
and hides it on Ready — so cache hits never flash it and serverless mode never
shows it.
Layer separation (per direction): ImplantPending belongs to the diegetic
implant UI only. The global, non-diegetic UI layer must use its own busy
indicator with its own visuals — sharing the sweep logic is fine, resting on
this theme/these glyphs is not. Documented in the component header.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
show_body() now requests the body's Layer-1 cascade output via
SimBridge.request_atlas_layers(); the response feeds set_generation_layer1().
The proxy returns Pending on a cache miss and generates in the background
(D-225), so the viewer re-requests every 0.5s (20-retry ceiling) until Ready,
guarding against stale responses by body_id. Serverless/test mode is a no-op
(overlays stay empty). Connected in _ready, torn down in _exit_tree.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Formatting-only cleanup of the #969/#960-A codec + bridge scripts to match
gdformat output (the pre-push gdformat check is advisory; these landed
un-formatted). No behavior change — 70/70 protocol tests green.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Render the proxied Layer1Output (D-225) as toggleable Atlas overlays. The
viewer stores the decoded Layer1Output via set_generation_layer1() and the
marker overlay draws three new layers gated by their visibility flags:
RVR river_cells as dots, confluences as small circles, mouths as
double-ring sea-terminus markers
BAS drainage-basin boundaries as thin closed polylines + faint fill
ATR geographic attractors — shape by attractor_type (Araminta's 7-shape
vocabulary), color by sub_biome, size by strength; <0.15 culled
Layer-1 positions are [row, col] in the 512x256 working grid, which matches
the viewer's grid_to_canvas transform, so they project directly onto the
heightmap. OVERLAY_DEFS gains gen_l1_rivers/basins/attractors (toggle group)
so AtlasOverlayBar auto-exposes them as toggle buttons.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Client transport half of the layer-stream protocol.
- SimBridge.request_atlas_layers(body_id) sends an AtlasLayerRequest frame
(live mode only; no-op in test mode); responses arrive via a new
atlas_layers_received signal.
- receive_bytes now decodes each frame ONCE via Protocol.decode_inbound and
branches by shape (snapshot vs atlas response) — avoids double-decoding the
20 Hz snapshot path. decode_snapshot is split into decode_raw +
_decode_snapshot_from_raw (public decode_snapshot unchanged, so the 70 protocol
tests stay the regression guard); decode_inbound returns {kind, value}.
70/70 protocol tests pass, including the new decode_inbound classifier test.
(Pre-existing client-suite failures in server-dependent e2e/roundtrip + unrelated
audio/fog/dialogue suites are unchanged — verified identical at baseline.)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Client half of the layer-stream protocol (codec only; transport wiring next).
- protocol.gd: encode_atlas_layer_request (bare {body_id, up_to} map so the
server demux routes it to the proxy, not the PlayerInput array) and
decode_atlas_layer_response (-> {body_id, status, error, layer1}; returns null
for non-atlas frames, e.g. a snapshot, so receive_bytes can disambiguate).
- gen_fixtures.rs: cross-language fixtures (atlas_response_ready/pending/
not_found) from real rmp_serde output, matching the test_protocol.gd pattern.
- test_protocol.gd: 5 tests decode the fixtures + verify a snapshot is not
mistaken for a response + the request encodes to the right shape. 68/68 pass.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closes the server-side layer-stream loop. serve_atlas_requests (PreInput) drains
the AtlasRequestBuffer and runs each through handle_atlas_request (cache hit ->
Ready; miss -> resolve via BodySourceResolver + enqueue an Immediate AnalyzeBody
-> Pending), buffering AtlasLayerResponses. send_atlas_responses (PostSnapshot)
flushes them to the client. main.rs wires BodySourceResolverResource (base root
= systems.db's 3rd ancestor; mod roots layer on later). Misses flow through the
#968 background tier and a re-request hits the now-warm cache.
Full path now live server-side: client request -> receive() demux -> serve ->
proxy -> (cache | queue+cascade) -> response -> client. The client half (send
request, decode response, render overlays) is #960.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Replaces the fixed-type SimBridge::receive_inputs() with a tagged
receive() -> Option<Inbound>, where Inbound is Inputs(Vec<PlayerInput>) or
AtlasRequest(AtlasLayerRequest). A shared decode_inbound() demuxes a frame by
shape (msgpack array = inputs, map = atlas request) — additive, no wire change
to existing input/snapshot frames. Adds send_atlas_response() to the trait
(both TcpBridge + LocalBridge impls). receive_bridge_inputs routes inputs to
the InputQueue as before; atlas requests to a new AtlasRequestBuffer (drained
by the serve system next). Integration tests (bridge_tcp/bridge_ipc) updated to
the tagged receive(); a demux unit test covers all three branches.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
handle_atlas_request: cache hit -> serialize Layer1Output, reply Ready; miss ->
resolve the source heightmap (BodySourceResolver) + enqueue an Immediate
AnalyzeBody on the #968 background queue, reply Pending (client re-requests; the
drain system populates the cache so a later request hits); unknown/no-terrain ->
NotFound. Adds the AtlasLayerRequest/Response/Status wire types (+ Serialize on
CascadeLayer). Pure handler; the bridge routing is the proxy's other half.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
First piece of the layer-stream proxy. Resolves body_id -> source heightmap.png
read-only from systems.db (terrain_reference, mirroring CultureResolver's
pattern), searching roots mod-first over the base install. Explicit errors
(UnknownBody / NoTerrainReference / SourceMissing). v1 callers pass the base
root only; the search-order logic is proven with synthetic mod roots so the
mod-first seam is ready (Q-099 covers registering a mod's new bodies).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
A multi-layer debug/review harness that reuses the real client UI so a human and
an automated agent can inspect deterministic server-computed state by attaching
to the live (auto-pausing) server. Supersedes the offline file-dump idea from
the D-225 discussion — the review tool is the production tool, no divergence,
no stale dumps.
Stack (bottom-up): live-pause substrate (TickRate::Paused + paused-allowlist,
triggered by gameplay_occluded; freezes world phases, keeps the bridge/gen-drain
alive) -> layer-stream proxy (D-225) -> human-visual viewer (#960, additive
overlays; shape=attractor type, color=sub-biome) -> agent channel
(AtlasAgentInterface: JSON observe + named-intent act, headless) -> interactive
capture (reuses the existing tests/run-visual primitive; complements, does not
supersede, the visual-golden regression role).
Consumers: Layer-1 geography now; economics + save-state inspection adopt the
pattern in their phases. Designed with Tyre (channel/pause/headless) + Araminta
(overlay encoding + affordance UX). Rides existing seams — a naming-and-contract
exercise, not a new subsystem.
Tickets: #970 auto-pause, #969 proxy, #960 viewer, #971 agent channel,
#972 capture (build order bottom-up).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The D-206 queue (#924) and D-203 cache (#917) were built but never connected to
the running app — the whole background tier was inert. This closes the loop:
submit -> Rayon -> cascade -> completion -> drain -> cache.
- gen_queue.rs: GenWorkItem::AnalyzeBody now carries enqueuer-resolved inputs
{ body_id, heightmap_path, sea_level, body_seed: SeedChain } (D-225 boundary —
run_work_item stays pure compute, no path/DB resolution). run_work_item runs
the real cascade: load heightmap.png -> downsample to GRID_W×GRID_H working
grid (D-202) -> run_layer1 -> BodyWorldState; load failure -> Failed.
GenCompletion::BodyAnalyzed carries the computed BodyWorldState.
- cascade.rs: CascadeSnapshot::into_body_world_state() conversion.
- plugin.rs (new): GenerationPlugin registers GenerationQueue +
BodyWorldStateCache and adds a PreInput drain system that inserts BodyAnalyzed
states into the cache (off the Rayon workers — a cheap channel drain, never
the ~45ms cascade). Wired into both the production and test app setups.
Tests run the real cascade on a tiny temp heightmap PNG (no committed fixture);
the plugin test proves the full submit->...->cache loop. The proxy (#969) is the
production submitter.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Resolves how per-body cascade layer data reaches the Godot Atlas viewer (#960),
on the lead's mod-first directive + Tyre's design pass:
- No bake (privileges first-party content + ~100MB install bloat). A server-side
layer-stream proxy computes on demand from moddable source files and streams
Layer1Output to the client.
- Transport: additive message tag on the existing IPC stream (no second socket);
single framed MessagePack response; raster stays a disk load.
- Mod-first BodySourceResolver: mod dirs override base install; heightmap.png is
the sole source of truth.
- Cache via the D-206 background queue on miss (never blocks the tick thread,
D-203); eviction -> recompute (deterministic, ~45ms).
- Whole Layer1Output per response; client composites additive overlays.
Q-098 resolved against its own premise (no durable store needed — recompute on
eviction is fine for a viewer). Q-099 spun off: mods adding new bodies need
terrain_reference rows in the binary systems.db (D-189), out of scope for #960.
Implementation decomposed into #968 (activate AnalyzeBody cascade run + cache
populate — the long pole) -> #969 (proxy + bridge protocol) -> #960 (viewer).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
- Pin the Body domain id scheme: body_id(String) -> FNV-1a(64) -> derive(Body, …),
via SeedChain::for_body. Closes the gap where "keyed by body StableId" named a
numeric id that doesn't exist (bodies are strings).
- Add for_body to the contract; document the stability guards (SeedDomain
#[repr(u64)] + pin test, AttractorType #[repr(u8)]).
- Note EntityRng keeps its domainless combine (re-expressing as derive(Npc, …)
is a stream-changing migration, deferred).
- Correct the implementation note: golden is JSON (matching golden_suite.rs),
not msgpack; run at 256x128 for a non-empty river network. Fix the stale
skeleton_gen.rs:248 line reference.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Addresses the Hoshe (QA) + Tyre (architecture) review of the SeedChain/cascade
work:
- Golden was pinning an empty river network (128x64 produced 0 river cells).
Bumped to 256x128, where GJ1c yields a real network (93 river cells, 19
mouths) — Layer 1's rivers are now actually guarded, not just attractors.
- SeedChain::for_body(world_seed, body_id) + fnv1a_64: the single canonical
body_id(String) -> u64 path (FNV-1a, the repo convention), so callers can't
derive divergent worlds from the same seed via different ad-hoc hashes. The
golden now uses it.
- Stability guards: seed_domain_discriminants_are_pinned test (CI fails if a
SeedDomain tag is renumbered); AttractorType gains #[repr(u8)] + explicit
discriminants (it's cast as a sort key in features.rs).
- Tests: SeedChain::root(0) non-degenerate; run_cascade error path (missing
file -> Err, not panic).
- Comments: clarified the id=0 derivations (sibling separation is caller-side
via the per-district/quarter chain; #957 threads the index), tightened the
all_district_types reachability comment (it pins the seed-0 sequence, not a
probabilistic claim), and noted the golden's WORLD_SEED is cosmetic at
Layers 0-1 + the x86_64 f32 capture caveat.
Deferred with reason: the run_cascade -> CascadeInputs struct refactor (Tyre)
is left for #954 — designing Layer-2's context shape now would be later-phase
detail, and there's a single caller to migrate then. EntityRng keeps its
domainless combine (migrating is stream-changing) — noted in D-224.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
End-to-end determinism guard. cascade_golden.rs pins two artifacts for a real
committed body heightmap (GJ1c) in one diffable JSON golden
(server/tests/golden/cascade_layer1.json):
- Layer 0: SHA-256 of the source heightmap.png bytes (flips if the Python
heightmap generator or the file changes)
- Layer 1: the serialized Layer1Output of run_cascade on a 128x64 downsample
(flips if the Rust drainage/feature/sub-biome code changes)
JSON (not the msgpack discussed in refinement) to match the existing
golden_suite.rs convention and stay diffable — a failure shows what drifted.
UPDATE_GOLDEN=1 regenerates; wired into `make golden-update`.
Adds Serialize/Deserialize to RiverNetwork/DrainageBasin/Layer1Output and a
sha2 dev-dependency.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
