test(api): cover functions the process-global mock constraint left untested

[GravityDarkLab]

What changed

Unit tests for functions that were previously untestable because Bun's mock.module() was process-global (route tests stubbed their modules for every file). Now that the API suite runs with --isolate (#10), each of those gaps is closed and its NOTE comment removed:

File	Newly covered
engine.test.ts	getCandidates, runFullMatchingPass — eligibility, active-contact exclusion, ranking + topN, rerank fallback and re-sort, 10-candidate cap, and per-applicant degradation (missing embedding / throwing score → that applicant gets [], the pass continues, including across concurrency batches)
match-state.service.test.ts	promoteAppliedToMatched — portal-score floor, participant dedup, only promotes from applied
admin.service.test.ts	listApplicants (filter, paging, magicToken/passwordHash never returned), deactivateApplicant (soft-delete write, malformed id, no match)
ai.service.test.ts	generateChatCompletion with fetch stubbed — request shape, and every failure mode returns "" instead of throwing
proposals.test.ts	stale import-path note removed (nothing was untested)

Why

runFullMatchingPass's per-applicant degradation shipped in #10 without a unit test, specifically because of this constraint. The engine tests now pin it: a mutation that removes the per-applicant catch fails two of them.

getOrComputeEmbeddings isn't covered here on purpose — #22 adds its tests alongside a fix to its fallback semantics.

Testing

• bun run typecheck clean
• bun run test:api: 609 pass / 0 fail (was 572; +37)
• Stable under bun test --isolate --randomize across 14 seeds

🤖 Generated with Claude Code

[Copilot]

🟡 Changes recommended

Two moderate coverage gaps and one tautological threshold assertion remain.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

Adds isolated API unit-test coverage for matching, applicant administration, match promotion, and AI completion fallbacks.

Changes:

• Covers matching eligibility, ranking, degradation, and batching.
• Tests applicant listing/deactivation, promotion rules, and AI failure handling.
• Removes obsolete test-isolation notes.

File summaries

File	Reviewed changes
api/src/__tests__/unit/matching/engine.test.ts	Adds candidate and full-pass matching tests. Moderate finding (3 votes): test doubles should enforce status predicates and cover inactive applicants.
api/src/__tests__/unit/services/ai.service.test.ts	Tests chat completion requests and fallbacks. Moderate finding (2 votes): add malformed-body or rejected-JSON coverage.
api/src/__tests__/unit/services/admin.service.test.ts	Tests applicant listing, secret omission, pagination, and deactivation.
api/src/__tests__/unit/matching/proposals.test.ts	Removes an obsolete mock-isolation note.
api/src/__tests__/unit/services/match-state.service.test.ts	Tests applicant promotion rules. Nit (1 vote): pin the documented 0.6 threshold rather than importing it from the implementation.

Review details

Suppressed comments (1)

api/src/tests/unit/services/match-state.service.test.ts:715

• The expected threshold is imported from the same implementation under test, so this assertion is tautological: changing PORTAL_MIN_SCORE and the query together would still pass. Pin the documented 0.6 floor (or assert the exported constant separately) to make this regression check meaningful.

    expect(filter).toEqual({ status: "proposed", score: { $gte: PORTAL_MIN_SCORE } });

• Files reviewed: 5/5 changed files
• Comments generated: 2
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.