## Database Migration Fix Applied ### Problem The /api/usage/breakdown endpoint was returning HTTP 500 errors due to missing database column `is_research_report` in the `queries` table. ### Root Cause The database was created outside of alembic (using SQLAlchemy directly), so the `alembic_version` table was empty. When `alembic upgrade head` was run, it tried to run ALL migrations from scratch, which failed because the tables already existed. ### Solution Applied 1. Stamped the database to the initial migration: `alembic stamp cc9d2f3bc955` 2. Ran remaining migrations: `alembic upgrade head` This applied the following migrations: - a534488c3a48: add cost breakdown columns to api_usage_log - b8c21e4f5d67: add missing API fields to search_results and images - c9d3e4f6a789: add source field to search_results and title field to images - d1e2f3a4b5c6: add async_queries table for deep research mode - e2f3a4b5c6d7: add is_research_report flag to queries table (the critical one) - f3a4b5c6d7e8: add progress tracking fields to async_queries table ### Verification - All 414 backend tests pass - /api/usage/breakdown endpoint now returns valid JSON responses - /api/usage/summary endpoint continues to work ### Note for Future Deployments When setting up a new database or environment, always run `alembic upgrade head` after creating the database to ensure all schema migrations are applied.

[QA] FAILED - Database migration required ## Summary The code implementation and tests are excellent (414/414 tests pass), but the /api/usage/breakdown endpoint is non-functional in production due to a missing database column. ## Critical Issue - **Error**: sqlite3.OperationalError: no such column: queries.is_research_report - **Cause**: Production database missing column added in migration e2f3a4b5c6d7 - **Impact**: All requests to /api/usage/breakdown return HTTP 500 ## Required Fix ```bash cd backend alembic upgrade head # Then verify: sqlite3 perplexity_clone.db ".schema queries" | grep is_research_report ``` ## Test Results - ✅ All 414 automated tests pass - ✅ 35 usage endpoint tests pass (18 summary + 17 breakdown) - ❌ Manual testing fails - endpoint non-functional - ❌ Database schema doesn't match models ## Details See full QA report: /tmp/qa-PPLX-44.md The implementation agent did excellent work. This is a deployment/migration issue, not a code quality issue.

## Implementation Complete ### Summary Added comprehensive tests for usage endpoints including a new breakdown endpoint. ### Changes Made **1. New /api/usage/breakdown endpoint** (backend/routers/usage.py:116-235) - GET endpoint with pagination (page, page_size) - Sorting by date, cost, or tokens (asc/desc) - Filtering by model, thread, date range - Returns individual usage entries with query details **2. CRUD function** (backend/crud.py:830-945) - get_usage_breakdown() with full filter/sort/pagination support - Properly handles null costs and tokens - Limits page_size to max 100 **3. Schemas** (backend/schemas.py:353-373) - UsageBreakdownEntry: Individual usage entry - UsageBreakdownResponse: Paginated response **4. Tests** (backend/tests/test_usage.py:666-1058) Added 17 new tests for breakdown endpoint covering pagination, sorting, filtering, error handling, and authentication. ### Verification - All 414 backend tests pass - All 35 usage endpoint tests pass - Branch merged to main and pushed to remote