fix(llm): a $0 catalog row silently disables the cost cap — plus the models.dev sync it unblocked

apps/cli/src/engine/catalog-metadata.test.ts

@@ -124,8 +124,20 @@ describe('catalog-metadata (ADR-0072 P4 — host projection + seed + install + p
 
     it('a reseed CARRIES FORWARD enrichment a refresh had populated — never blanks it (ADR-0072 point 5)', () => {
       seedShippedCatalog(store, TS);
-      const shippedId = Object.keys(CATALOG_SNAPSHOT)[0]!;
-      // A `models refresh --catalog` populated enrichment on a shipped id (the snapshot itself carries none).
+      // Pick a shipped id the snapshot leaves WITHOUT a `knowledgeCutoff`, so this test exercises the
+      // carry-forward branch rather than the snapshot-wins one. The snapshot bakes in modalities and a
+      // description for every model today, but not always a cutoff.
+      const withoutCutoff = Object.entries(CATALOG_SNAPSHOT).find(
+        ([, m]) => m.knowledgeCutoff === undefined,
+      );
+      // If a future snapshot stamps a cutoff on EVERY model, this test can no longer reach the branch it
+      // exists for — say so, rather than dying on an unhelpful undefined-index read.
+      expect(
+        withoutCutoff,
+        'no shipped model lacks knowledgeCutoff — pick another carry-forward column',
+      ).toBeDefined();
+      const shippedId = withoutCutoff![0];
+      // A `models refresh --catalog` populated enrichment on a shipped id.
       store.updateEnrichment([
         {
           modelId: shippedId,
@@ -139,10 +151,13 @@ describe('catalog-metadata (ADR-0072 P4 — host projection + seed + install + p
       store.upsertMeta({ seededSnapshotSha: 'a-different-sha' });
       expect(seedShippedCatalog(store, TS + 1)).toBe(true); // it reseeds
       const row = store.readAll().find((r) => r.modelId === shippedId)!;
-      // Money+wire came from the (re-reviewed) snapshot; enrichment SURVIVED the reseed rather than resetting to NULL.
+      // The column the snapshot leaves empty SURVIVES the reseed rather than resetting to NULL...
       expect(row.knowledgeCutoff).toBe('2099-01');
-      expect(row.inputModalities).toBe(JSON.stringify(['text', 'image']));
-      expect(row.description).toBe('fetched');
+      // ...while a column the snapshot DOES carry wins over the stale DB value, which is the documented
+      // `?? prior` precedence: the newly-reviewed snapshot is the more trustworthy source for what it states.
+      const shipped = CATALOG_SNAPSHOT[shippedId]!;
+      expect(row.inputModalities).toBe(JSON.stringify(shipped.inputModalities));
+      expect(row.description).toBe(shipped.description);
     });
   });
 

docs/roadmap/current.md

@@ -337,9 +337,9 @@ unanswered**; the remainder sit inline in their own phase-file bullet.
 |---|------|----------|----------------|
 | D1 | 0 | Mark the `ci` job a **required** check (open since Phase 0) | Yes — every CI item here is advisory until it flips |
 | D2 | 0 | Coverage floor: promote to required, or soften `testing.md`? | Promote, **and implement in the same PR** — a checked-in run already shows 92–97% margin |
-| D3 | 0 | Accept the upstream Gemini price rise the ADR-0071 §9 guard is refusing? | Accept and regenerate — a stale floor *under*-prices the cap, the dangerous direction |
+| D3 | 0 | Accept the upstream price changes the ADR-0071 §9 guard is refusing? | ✅ **Ruled 2026-07-26: take current prices.** Verified: `gemini-flash-latest` moved $0.30→$1.50 in / $2.50→$9.00 out — the shipped floor under-prices by 5×. **Blocked on D3b below** |
 | D4 | 0 | Confirm the number reservation (`0013`–`0017`, ADR-0073+) | As listed — one item per number, in landing order |
-| D5 | 0 | The binding locale bar: EXIT:6 says five, CLAUDE.md says `en`+`tr` | **Maintainer call.** Either amend EXIT:6 to ship `en`+`tr` with three staged, or keep five and fix the prose. 2.5.5.F assumes the latter |
+| D5 | 0 | The binding locale bar | ✅ **Ruled 2026-07-26: ship `en` + `tr`**, catalog architected for n locales, `es`/`fr`/`de` staged. EXIT:6, 2.6.L and 2.5.5.F amended |
 | D7 | 0 | Publish v0.1.1 as-is, or supersede with v0.2.0? | v0.2.0 — ADR-0067's Node `>=22` bump is breaking for 0.x |
 | D8 | 1 | Do already-persisted approval previews need a scrub? | Yes — migration 0013, same PR. Deleting `history.db` also destroys provider registrations |
 | D10 | 1 | `BudgetExceededError`/`BudgetPauseError`: adopt `.code`? | Adopt — must precede Wave 3's `RelaviumError` migration |

docs/roadmap/phases/phase-2.5.5-hardening-and-remediation.md

@@ -345,7 +345,7 @@ The project's own documentation conventions — cite-not-restate, status banners
 - **Repo-wide docs link/index hygiene sweep: four broken relative links, three missing ADR cross-reference pointers, and four stale reference-doc index tables.** Superseded ADR-0021 has no in-body amendment pointer forward to the new Node floor decision. ADR-0069's Related line 404s to a renamed file — it links `0047-cli-render-seam-and-framework-free-cores.md` but the real file is `docs/decisions/0047-cli-framework-commander-ink-clack.md`. ADR-0071's `### K7.` header gives no in-header signal that it belongs to a separate numbering track. An independent repo-wide crawl of every relative markdown link under `docs/` found four genuinely broken links — one of which duplicates the ADR-0069→ADR-0047 link above — plus a wrong-subdirectory link in ADR-0062. Separately, `docs/reference/README.md`'s shared-core table lists only 5 of 11 real files; `docs/standards/documentation-style.md` §6's canonical-artifact registry is missing 6 of 11, including the safety-critical `llm-provider-seam.md`; `docs/reference/shared-core/agent-runner.md` is orphaned from both indexes despite being cited by six other docs; the `docs/reference/cli/` index is missing `accessibility.md` and `regression-harness.md`. Regenerate all four index tables from the real file tree, fix the four broken links, and add the three missing ADR pointers in one coordinated sweep; recommend a CI link-checker afterward to prevent recurrence. *(M · docs/decisions/, docs/reference/README.md, docs/standards/documentation-style.md, docs/reference/cli/ · #123, #124, #126, #143, #151, #251, #252, #253)*
 - **Trim two ADRs and one project-structure.md row that restate content already living verbatim in their own cited canonical doc.** ADR-0031 states "canonical types live in `llm-provider-seam.md`" and then spends roughly 250 lines re-deriving the exact same content that file already carries. ADR-0027's addendum states exact sandbox resource-cap numbers that already live in `expression-sandbox-spec.md`, even after the ADR's own preamble names that spec the single canonical home. `docs/project-structure.md`'s Apps table restates a hand-picked, now-stale CLI subcommand list instead of linking to `docs/reference/cli/commands.md`, which already owns it canonically, and has drifted out of sync as a direct result. All three are the identical CLAUDE.md rule-8 violation — trim each ADR to drivers/tradeoffs plus a citation, and replace the stale restated list with a link. *(M · docs/decisions/0031-*.md, docs/decisions/0027-*.md, docs/project-structure.md · #120, #121, #257)*
 - **Add three missing rows/notes to `tech-stack.md`, the single source of truth for pinned choices.** `docs/tech-stack.md` omits `smol-toml` and `string-width` as ADR-backed runtime dependencies (both already shipped and in `package.json`), and is missing an MCP SDK shape-tradeoff note that should cross-reference ADR-0034. The Playwright e2e row is the one un-flagged not-yet-applicable row in an otherwise carefully phase-annotated doc. Add the two missing dependency rows, the tradeoff note, and the same Phase-3 flag every other not-yet-applicable row already carries, in one pass. *(S · docs/tech-stack.md · #246, #247, #140)*
-- **Rewrite CLAUDE.md's and README.md's shared i18n-status clause: wrong on stage, wrong on language count, repeated in both files.** `CLAUDE.md`'s "What this project is" section lists "settings/theming/`en`+`tr` localization" as in-progress Phase 2.6 scope, reading as already shipping — an exhaustive grep for i18n/locale/translation scaffolding across `apps/cli` finds zero hits; the theme hasn't started at all. The same clause also undercounts scope: `docs/roadmap/phases/phase-2.6-conversational-authoring.md` states the real target three times as `en, es, tr, fr, de`, and `README.md` repeats the same en/tr undercount. One rewrite of the shared clause in both `CLAUDE.md` and `README.md`, stating the correct five-language target with an explicit not-started qualifier, fixes all three. *(S · CLAUDE.md, README.md · #74, #260, #134)*
+- **Rewrite CLAUDE.md's and README.md's shared i18n-status clause: wrong on stage, wrong on language count, repeated in both files.** `CLAUDE.md`'s "What this project is" section lists "settings/theming/`en`+`tr` localization" as in-progress Phase 2.6 scope, reading as already shipping — an exhaustive grep for i18n/locale/translation scaffolding across `apps/cli` finds zero hits; the theme hasn't started at all. The language count is now settled by maintainer ruling (2026-07-26): the catalog is architected for n locales but **`en` + `tr` are what ship in Phase 2.6**, with `es`/`fr`/`de` staged — so the existing en/tr wording is correct on scope and wrong only on stage. One rewrite of the shared clause in both `CLAUDE.md` and `README.md`, adding an explicit not-started qualifier and naming the staged locales, fixes all three. *(S · CLAUDE.md, README.md · #74, #260, #134)*
 - **Add the six load-bearing terms missing from the glossary and refresh its stale "Last updated" header.** `docs/glossary.md` omits `ExecutionHost`, `ToolHost`, `ActionGuard`, `MediaStore`, `WorkflowEngine`, and `BudgetGovernor` — terms that recur across the ADR corpus without a defined home — and its "Last updated" header is over a month stale relative to the file's own content. Bundle both fixes into the same edit. *(S · docs/glossary.md · #135, #137)*
 - **Swap `docs/architecture/README.md`'s recommended reading order to match both target docs' own stated shipping order.** The recommended order puts `cloud-phase-2.md` before `managed-inference.md`, contradicting both documents' own stated shipping/dependency order (managed inference is Phase 5, cloud execution is Phase 6). Swap the two entries. *(S · docs/architecture/README.md · #157)*
 - **Add the missing `style` commit type to `commit-style.md`'s enumerated Types list.** `docs/standards/commit-style.md` enumerates `feat`, `fix`, `refactor`, `perf`, `test`, `docs`, `chore`, `build`, `ci` but omits `style`, which real repo history already uses (e.g. `style: prettier the ADR-0072 files`, HEAD-2). Add it to the documented list so the doc matches actual practice. *(S · docs/standards/commit-style.md · #158)*
@@ -354,7 +354,7 @@ The project's own documentation conventions — cite-not-restate, status banners
 - **Correct the doc comment on `urlSourceSchema` that asserts the opposite of the flag's actual current state.** A doc comment attached to `urlSourceSchema` (a security-relevant media feature flag) states the flag's behavior backwards from what the code actually does today. Correct the comment to match the real, current gating. *(S · packages/shared/src/content.ts (or the file defining `urlSourceSchema`) · #238)*
 - **Sync `agent-runner.md`'s documented `cost:updated` payload with the `priced` boolean field ADR-0070 added.** `docs/reference/shared-core/agent-runner.md` predates ADR-0070 and its documented `cost:updated` event payload is missing the `priced` boolean field ADR-0070 added to `CostUpdatedEventSchema`. Add the field to the documented shape and cite ADR-0070 as the source. *(S · docs/reference/shared-core/agent-runner.md · #145)*
 
-**Acceptance:** no canonical doc names the retired `pricing.ts` `MODEL_PRICING` table as the pricing authority; every reference to an unbuilt surface (desktop, VS Code, `apps/api`/`portal`, `packages/ui`) carries the same not-yet-shipped marker convention already used correctly elsewhere in the corpus, at all twenty flagged sites; every reference-doc index (`docs/reference/README.md`, `docs/reference/shared-core/README.md`, `documentation-style.md` §6, `docs/reference/cli/` index) lists every real file in its directory with zero broken relative links, verified by a link-check pass; `packages/mcp` appears in all five package-inventory tables/diagrams including the reviewer agent's own checklist; every roadmap status banner (`current.md`, `AGENTS.md`, the phase-2.6 and phase-3 banners) matches the state recorded in the file it summarizes, with the 2.6.Q "seven decisions" figure corrected to six; the chat tutorial and onboarding path reflect that `chat.ts` ships today; `node-types.md`'s `merge_strategy`/`human_in_the_loop_config` claims match the schema and engine exactly; the three hardcoded-path skills run correctly from a fresh clone via `` `$(git rev-parse --show-toplevel)` ``; `commit-style.md` lists `style` as a valid type; the glossary defines all six missing terms and its date header is current; CLAUDE.md and README.md state the i18n theme as not-started with the correct five-language (`en, es, tr, fr, de`) target. The two blocked items (`overview.md`'s LLM→Keychain edge, Home's `showCost` wiring) surface as explicit open questions in the phase file rather than being silently resolved by guess.
+**Acceptance:** no canonical doc names the retired `pricing.ts` `MODEL_PRICING` table as the pricing authority; every reference to an unbuilt surface (desktop, VS Code, `apps/api`/`portal`, `packages/ui`) carries the same not-yet-shipped marker convention already used correctly elsewhere in the corpus, at all twenty flagged sites; every reference-doc index (`docs/reference/README.md`, `docs/reference/shared-core/README.md`, `documentation-style.md` §6, `docs/reference/cli/` index) lists every real file in its directory with zero broken relative links, verified by a link-check pass; `packages/mcp` appears in all five package-inventory tables/diagrams including the reviewer agent's own checklist; every roadmap status banner (`current.md`, `AGENTS.md`, the phase-2.6 and phase-3 banners) matches the state recorded in the file it summarizes, with the 2.6.Q "seven decisions" figure corrected to six; the chat tutorial and onboarding path reflect that `chat.ts` ships today; `node-types.md`'s `merge_strategy`/`human_in_the_loop_config` claims match the schema and engine exactly; the three hardcoded-path skills run correctly from a fresh clone via `` `$(git rev-parse --show-toplevel)` ``; `commit-style.md` lists `style` as a valid type; the glossary defines all six missing terms and its date header is current; CLAUDE.md and README.md state the i18n theme as not-started, shipping `en` + `tr` with `es`/`fr`/`de` staged (the 2026-07-26 ruling). The two blocked items (`overview.md`'s LLM→Keychain edge, Home's `showCost` wiring) surface as explicit open questions in the phase file rather than being silently resolved by guess.
 
 ### 2.5.5.G — Codebase hygiene: naming, duplication & dead code