Analyzes all 93 existing backend tests and identifies critical gaps:
- Agent service (1,074 lines, 0% coverage)
- Upload review/approve workflow (6 untested endpoints)
- Utility modules (questions.py, csv_files.py — 0% coverage)
- Frontend has no test framework configured

https://claude.ai/code/session_01DDzfevLjm4dzjb7qxWAM98

- test_questions.py (22 tests): clean_display_question and normalize_question_key
  for question text normalization and deduplication key generation
- test_csv_files.py (14 tests): CSV format detection, round-trip read/write,
  BOM handling, delimiter preservation, special characters
- test_agent_unit.py (66 tests): agent mode normalization, provider aliases,
  availability checks, trace logging, JSON extraction from LLM output,
  retry/backoff logic, error detail extraction, response retriability
- test_review_workflow.py (17 tests): question result listing, editing,
  approval, bulk approval, download guards, and job finalization

https://claude.ai/code/session_01DDzfevLjm4dzjb7qxWAM98

- test_export_and_flagged.py (14 tests): QA export as CSV/JSON with
  category filtering, empty KB edge case; flagged question get/dismiss
  with duplicate group resolution; deduplication endpoint
- test_upload_helpers.py (17 tests): output file spec generation,
  media type mapping, form value cleaning, RunFormat serialization
  round-trips

Total new tests across both commits: 150

https://claude.ai/code/session_01DDzfevLjm4dzjb7qxWAM98

- test_parser_helpers.py (31 tests): confidence scoring, fallback
  decision logic, text normalization, header pattern matching, and
  table column mapping inference for multi-column layouts
- test_database.py (4 tests): schema migration idempotency, column
  addition from minimal schema, and fresh-schema no-op verification
- Updated TEST_COVERAGE_ANALYSIS.md to reflect new totals (278 tests)

https://claude.ai/code/session_01DDzfevLjm4dzjb7qxWAM98