bird-agents: pre-copy template into per-task working sqlite (DEV-1509)

[ZmeiGorynych]

Summary

• Bug. Cloud-run agent saw OperationalError: no such table: artifactscore from the dry-run gate (agents/_submit.py::_dry_run_sql) and burned 40-100 turns rationalising it as a casing problem. The Linear issue (DEV-1509) framed it as a SLayer-introspection-vs-SQLite casing mismatch; investigation showed that framing is wrong.
• Actual cause. harness.materialize_task_db only symlinked the template at <id>/<db>_template.sqlite and left the working <id>/<db>.sqlite uncreated. Anything that touched the path RW before the upstream's first reset_and_restore_database (most likely upstream get_db_connection's default-RW sqlite3.connect, which creates the file if missing) left an empty SQLite at db_file_path. The gate's sqlite3.connect("file:<db_file_path>?mode=ro", uri=True) then produced precisely no such table: <name> for every referenced table — empirically reproduced.
• Fix. materialize_task_db now pre-copies the canonical template to the working <db>.sqlite via shutil.copy2 + os.replace (atomic). The per-task <db>_template.sqlite entry remains a SYMLINK (DEV-1462 storage invariant: 18 templates × N tasks would blow storage). The fast-path idempotence guard additionally checks the SQLite magic header (catches 0-byte AND foreign shapes like LFS pointers). Both the symlink and working-file installs use per-call uuid-suffixed tmp paths + os.replace, so two concurrent materialize_task_db calls on the same instance_id cannot collide.
• Out of scope. The Linear issue's "third AC bullet" (a casing-regression test on ensure_db_cache) is intentionally skipped — slayer ingest already preserves SQLite-stored casing verbatim and SQLite identifier lookup is case-insensitive. A new test test_slayer_ingest_preserves_table_casing_and_sqlite_case_insensitive pins both invariants as documentation-via-code so future readers understand why the casing framing was rejected.

Tests

15 new tests in tests/test_db_isolation_pre_copy.py. All 9 existing tests/test_db_isolation.py tests still pass. Full non-integration suite: 1514 passed, 94 skipped.

Coverage map:

#	Behavior pinned
1, 2	Working file present + byte-equal to template after materialize
3, 4	_dry_run_sql + sqlite3.connect(?mode=ro) succeed against the materialised file
5	Template entry remains a symlink
6	Idempotent fast-path takes no copy when intact
7-9	Rebuild when working file missing / 0-byte / LFS pointer
10, 11	Stale-symlink rebuilds (with and without db_file_path set)
12	Upstream reset_and_restore_database still works after pre-copy
13	Documentation-via-code: slayer ingest preserves casing AND SQLite is case-insensitive
14	Both symlink AND working-file installs use atomic tmp + os.replace
15	Mid-copy shutil.copy2 failure leaves expected_db_file untouched

Test plan

• pytest tests/test_db_isolation_pre_copy.py tests/test_db_isolation.py -v — 24/24 pass
• Full non-integration suite: env -u SSH_AUTH_SOCK uv run --extra all --extra dev --extra pydantic-ai pytest — 1514 passed
• Manual cloud smoke (optional): rerun museum_5 / museum_7 with the same Opus/Sonnet config from 20260530t0958-claudes-slayer-b278f1 and confirm the casing rabbit hole does not reappear

🤖 Generated with Claude Code

[linear]

DEV-1509

[coderabbitai]

Warning

Review limit reached

@ZmeiGorynych, we couldn't start this review because you've reached your PR review rate limit.

More reviews will be available in 54 minutes and 18 seconds. Learn how PR review limits work.

Your organization has run out of usage credits. Purchase more in the billing tab.

⌛ How to resolve this issue?

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.

We recommend that you space out your commits to avoid hitting the rate limit.

🚦 How do rate limits work?

CodeRabbit enforces hourly rate limits for each developer per organization.

Our paid plans include higher PR review limits than trial, open-source, and free plans. In all cases, reviews become available again over time. During sustained high-volume PR review activity, CodeRabbit may temporarily slow when the next review becomes available.

Please see our Fair Usage Limits Policy for further information.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 683ac8f9-b1ec-4726-bc1a-6aa27c696ce6

📥 Commits

Reviewing files that changed from the base of the PR and between 0fb3ad5 and 9d4975f.

📒 Files selected for processing (2)

• src/bird_interact_agents/harness.py
• tests/test_db_isolation_pre_copy.py

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch egor/dev-1509-slayer-introspection-submission-table-name-casing-mismatch

---

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

[ZmeiGorynych]

@coderabbitai review now!

[coderabbitai]

@ZmeiGorynych Sure, I'll review the changes now!

✅ Actions performed

Full review triggered.