test(fixtures): restore sessions-a-few fixture with assertions

[pepmach]

Summary

Restores the sessions-a-few seed fixture and gives it a consumer test. The fixture seeds four chat sessions covering every session record state the dashboard distinguishes — pinned, ordinary open, closed-in-History, and archived — so each state is asserted to land in the correct view with the correct ordering rather than merely existing on disk.

Transcript JSONL is the seedable half of session state; the SQLite event log is not (per the SQLite deferral in rich/README.md), so nothing here depends on it — the test asserts no memory.db is created.

Backend-only: a test fixture directory plus its consumer test. No user-facing surface changes, so no screenshots.

Consumer test

test/test_fixture_sessions_a_few.py drives the production readers, not reimplementations of them:

• ConversationLog.list_sessions() — the session catalog, asserted to return exactly the three non-archived keys.
• The restore filter, _prefetch_recent_session() from kiro_crew.dashboard.chat_persistence — asserted to hydrate messages for the pinned and open records and to return (None, None) for the closed one, which is what keeps a closed thread in History instead of restoring it into a slot.
• ConversationLog.read_messages() — the transcript reader, asserted to yield the closed thread's user/assistant turns.
• backfill_transcripts() from kiro_crew.events.backfill — the archive backfill reader, asserted to emit the archived slice's two SessionMessage events with the expected roles and content lengths.

Deterministic os.utime mtimes (1_000.0 for pinned, 2_000_000_000.0 and +1 for open/closed) pin the ordering and the recency cutoff so the bucket assignment cannot drift with wall-clock time.

Schema repairs

The fixture was written against the current on-disk schema rather than the shape it originally carried:

• Archive header records use the current archive schema — _type: archive with reason, archived_at, and count.
• Session metadata records carry closed_at as an epoch float (1768467600.0), which is what the current readers parse; an ISO string in that field is not accepted by the closed-session path.

Verification

• .venv/bin/python -m pytest test/test_fixture_sessions_a_few.py test/test_pod_scenarios_command.py test/test_pod_seed_scenarios.py test/test_seed.py -q → 125 passed. The three sibling suites are the fixture registry's own coverage (test_pod_seed_scenarios.py walks every shipped fixture's manifest), so they confirm the new fixture registers cleanly alongside the registry-tolerant scenarios test from fix: return complete pod scenario descriptions #8222.
• Gates on the touched test file: isort --check-only, flake8, and black --check --target-version py310 all clean.
• Diff boundary: 7 files, +125/-0, confined to src/kiro_crew/tests_fixtures/sessions-a-few/ plus test/test_fixture_sessions_a_few.py.

Pattern harvest

Not generalizable: fixture restoration validated against current production readers; no new rule.

[github-actions]

GPT 5.6 Review — ✅ no blocking findings

GPT 5.6 completed its review of 87c2fba5c2449dcf0682bda5530d9b950fba18bb and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 87c2fba

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 87c2fba5c2449dcf0682bda5530d9b950fba18bb: <one-sentence reason>

[github-actions]

Design Review (Fable 5) — ✅ PASS

Design-level review of 87c2fba5c2449dcf0682bda5530d9b950fba18bb — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Test-only fixture restoration whose assertions ride the real readers (list, prefetch filter, transcript, backfill), each state verified against actual production filter semantics.

[DESIGN-REVIEWED] 87c2fba

[github-actions]

Opus 4.8 Review — ✅ no blocking findings

Reviewed 87c2fba5c2449dcf0682bda5530d9b950fba18bb — this comment is updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 87c2fba

Verdict parsed from the review's SHA-scoped output markers for commit 87c2fba5c2449dcf0682bda5530d9b950fba18bb.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 87c2fba5c2449dcf0682bda5530d9b950fba18bb: <one-sentence reason>

[github-actions]

First Principles Review (Fable 5) — 🟡 CONCERNS

Premise-level review of 87c2fba5c2449dcf0682bda5530d9b950fba18bb — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

I have what I need: the contract, the intent, the patch, the existing fixture inventory (empty/minimal/rich), the auto-discovery mechanism in seed.py, and the existing test coverage of the same readers. Emitting the review.

First-Principles-Verdict: CONCERNS

All four session states this fixture seeds already ship in rich, and the "backend-only" fixture auto-registers as a user-visible seed scenario.

What this change ships

Intent: give the dashboard's four session states a seeded fixture whose classification is asserted through the production readers — an ADDITION (+125/-0; "restore" names prior existence, not a defect).

1. A new seedable home sessions-a-few with one session per dashboard state — duplicate of rich's state coverage
2. The scenario appears in kirocrew pod scenarios and as a valid gateway --seed value — undeclared
3. A test asserting each seeded state lands in its dashboard bucket via the real readers — justified
4. Six fixture files ship in the runtime wheel to every install — rides along with 1

Watch

• Description claims "Backend-only … No user-facing surface changes", but available_fixtures() (src/kiro_crew/seed.py:103) discovers fixtures by directory listing, so this name joins the kirocrew pod scenarios output and the accepted --seed values — shipped surface that must be honored once released.
• Duplication, counted: grep pinned|closed over rich/sessions/ shows 4/4 states already present — 2 pinned, 1 open (triage-demo), 1 closed (research-notes), plus an archive/ slice. The new test sets its own os.utime mtimes, so nothing about it requires a fresh fixture.

Subtractions

• Delete src/kiro_crew/tests_fixtures/sessions-a-few/ (6 files) and point test/test_fixture_sessions_a_few.py at seeded_home("rich"), asserting rich's keys; rich's session set is already pinned by test/test_fixtures_helper.py::test_seeded_home_rich_has_multi_session_coverage, so the two tests would hold one contract instead of two fixtures diverging.

[FIRST-PRINCIPLES-REVIEWED] 87c2fba

[dwu96]

Tier 1 auto-approve: test (7 files). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, SAST annotations clean, security checklist all-NO, AI reviewers green. Category: purely additive test fixture — new tests_fixtures/sessions-a-few data tree plus one new test/ consumer test; no existing file modified and no runtime code path changed.