feat(filters): make formulation a real facet surface

[BigSimmo]

Summary

Step three of the filter contract rollout (docs/filter-contract.md), and the first facet adoption. Formulation rendered twelve domains as radios; both halves of that were wrong.

• Radios were the wrong claim. A mechanism carries 3.92 domains on average (measured over all 12), so one-of-N told the reader they could not hold Affect and Risk at once. domain is now a resultFilterFacetGroup — many-of-N, OR within the group — and searchFormulationMechanisms accepts a domains set alongside the older single domain that the builder page's own select still uses.

• Three of the twelve could never match anything.formulationDomains is the taxonomy; Biological, Social and Cultural are carried by 0 of the 12 mechanisms. That is the exact failure mode the contract's companion rule exists for: under union counting an empty option reports the unchanged total, so it looks identical to a full one. New formulationDomainsInUse derives the list from the mechanisms that carry it — 9, in taxonomy order — which is §3's "derive the option list from the data, never declare it".

• Counts use the same predicate as the filter, asked with the candidate added: "how many would I have if I ticked this as well". Verified monotonic, and verified non-additive — Affect (9) OR Risk (4) is 10, not 13, because the overlap is real. A zero that is a consequence of the current query renders as a focusable dead end carrying its reason, never a silent no-op — and never on an already-selected option, which would make a live constraint unremovable.

• The Pattern group was not a filter. It called router.push and replaced the query, discarding the search and its results with no warning and no undo — precisely what §1 says a control labelled "Filter" must not do. Those presets now render below the band as AnswerSuggestionChips, framed as a new search. All five, rather than the four the old .slice(0, 4) exposed; the fifth had been unreachable at every breakpoint.

• ResultFilterFacetChips is exported so the desktop rail renders the same control as the sheet. Lens modes get that from SegmentedControl; facets had no equivalent, and the alternative was a second hand-rolled chip row per mode — which is how the breakpoints came to disagree in the first place. The desktop <select> of 13 options is gone. resultFilterFacetGroup now returns the narrow ResultFilterFacetGroup rather than the union, so one group object can feed both the sheet (union) and the rail (facet-only).

No scope segment: 12 mechanisms total, so the two segments would show near-identical numbers. docs/filter-contract.md §4 gates it on services.

Ordering note

This shares result-filter-control.tsx with PR #1857 (the four lens modes), which is open and fixes the sheet's counted-option accessible name ("Attachment0" → "Attachment (1)"). I deliberately did not duplicate that fix here — it would conflict on identical lines. Land #1857 first; the fix then applies to these facet chips at both breakpoints for free, since they are now one shared renderer. If this lands first instead, #1857 rebases cleanly onto it.

Verification

• npm run verify:pr-local — failed: (none), not reached: (none). Every step green including build, run with the dev server stopped and .next deleted so the /issues#210 dev-types corruption could not reproduce:

- completed: check:runtime, check:installed-lock-parity, format:changed, sitemap:check, docs:check-index,
docs:check-inventory, docs:check-scripts, docs:check-links, check:branch-review-ledger,
check:outstanding-issues, lint, typecheck, test, build, check:rag:fixtures
- failed: (none)
- not reached: (none)

• Full offline unit suite — Test Files 558 passed (558), Tests 6101 passed | 4 skipped (6105), zero failures.
• tests/formulation.test.ts — 11 passed, +3 pinning the contract rules rather than the current numbers' shape:
  • only the 9 carried domains are offered, Biological/Social/Cultural excluded, every offered domain returns >0, taxonomy order preserved
  • union not intersection: Affect 9, Risk 4, both 10; both < affect + risk; an empty set imposes no constraint rather than matching nothing
  • every add-one-more count ≥ the current total, and re-offering a selected option reports the unchanged total
• npm run check:bundle-budget after rm -rf .next && npm run build, .next/BUILD_ID mtime checked against the build:

[bundle-budget] production (what users download, 101 routes): 1296.1 KiB gzip — baseline 1279.1 KiB, within tolerance.
[bundle-budget] mockups (design scratch, 404s in production, 68 routes, 79 exclusive chunks): 285.1 KiB gzip — baseline 267.5 KiB, within tolerance.

• npm run verify:ui when UI, routing, styling, browser behavior, reduced-motion, or forced-colors behavior changed — UI verification not run as the full Chromium gate, for the known Playwright/Chromium revision drift (/issues#255, PW 1.62.1 wants chromium-1234, only 1194 installed); forcing a mismatched binary is disallowed. Proved instead against the running dev server, 0 px horizontal overflow at 1440 / 390 / 320 px:
  • desktop rail shows 9 facet chips, not 12; <select> count on the page is 0; suggestion row present with 5 chips
  • live accumulation on ?q=thoughts: Behaviour starts a dead end (hint 0, aria-disabled=true), and after ticking Affect becomes hint 1 and selectable — the union widening, visible in the control
  • aria-pressed accumulates across Affect + Behaviour rather than replacing
  • phone sheet at 390 px: one group, radiogroup count 0 (no lens groups left), Pattern absent, dead ends carry "No matches with your current filters.", footer 1 showing

Not run, with reasons: eval:retrieval:quality / eval:rag — no retrieval, ranking or answer-path file is touched; searchFormulationMechanisms is static-content search over a 12-item bundle, not a RAG surface, and check:rag:fixtures passed inside the gate above (36 golden cases, 23 suites). verify:release — provider-backed, not authorised and not a release.

Risk and rollout

• Risk: Moderate, and confined to one mode plus one shared component. The behaviour change is real and intended — formulation's filter now selects differently (many-of-N), offers 9 options instead of 12, and no longer throws the query away. The shared-file edits are additive: ResultFilterFacetChips is the existing renderer exported under a name, with no change to what it renders, and resultFilterFacetGroup's return type narrowed to a subtype, so every existing call site still typechecks and behaves identically. searchFormulationMechanisms keeps its domain option untouched for the builder page.
• Rollback: revert 9262d89. Nothing else consumes formulationDomainsInUse, the domains search option, or the exported chips.
• Provider or production effects: None.

Notes

formulationDomains is deliberately left exported and unchanged — it is the taxonomy, and MechanismDomainChips and the builder still render against it. The derived list is a filter concern, not a redefinition of the domain set.

Next: services (facet index + selection state + URL + the scope segment — the biggest of the remaining, and where the scope segment finally earns its place), then factsheets, therapy-compass, documents last.

---

Generated by Claude Code

[supabase]

This pull request has been ignored for the connected project sjrfecxgysukkwxsowpy because there are no changes detected in supabase directory. You can change this behaviour in Project Integrations Settings ↗︎.

---

Preview Branches by Supabase.
Learn more about Supabase Branching ↗︎.

[coderabbitai]

Warning

Review limit reached

You’ve reached a temporary PR review limit under our Fair Usage Limits Policy.

Your recent review volume is higher than typical usage, so adaptive limits are currently applied.

Next review available in:51 minutes

Your organization has reached its usage spending cap. Adjust your spending cap in the billing tab.

How can I continue?

After more reviews become available, a review can be triggered using the @coderabbitai review command as a PR comment. Alternatively, push new commits to this PR.

To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews.

How do review limits work?

CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability.

For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window.

Please refer docs for additional details.

Review details

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: fc82f760-0668-427a-aa5c-45596169c2d5

📥 Commits

Reviewing files that changed from the base of the PR and between 2418a2a and deeab6a.

📒 Files selected for processing (7)

• docs/branch-review-ledger.md
• docs/outstanding-issues.md
• src/components/clinical-dashboard/result-filter-control.tsx
• src/components/formulation/formulation-home-page.tsx
• src/lib/formulation.ts
• tests/formulation.test.ts
• tests/ui-formulation.spec.ts

---

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

CI triage

CI failed on this PR. Automated classification of the 2 failed job(s):

• Production UI (1) — needs investigation: inspect the failing step and uploaded diagnostics; rerun only after classifying the cause.
• PR required — needs investigation: inspect the failing step and uploaded diagnostics; rerun only after classifying the cause.

Compared with main CI run #9928 (cancelled).

Classification is evidence routing, not permission to ignore a failure. Exact quarantined Playwright identities remain governed by the flake ledger.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit:9262d89701

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".