Add OpenAI API retry with exponential backoff for transient errors in /api/generate #629

New issue

Open

opened 2026-04-11 06:53:31 -04:00 by pook · 0 comments

pook commented

2026-04-11 06:53:31 -04:00

Owner

The /api/generate route calls OpenAI but has no retry logic — a single 429 or 500 fails the generation request.

Implementation:

Wrap the OpenAI API call in a retry helper: max 3 attempts, backoff 1s → 2s → 4s.
Only retry on 429 and 500/502/503. Do NOT retry on 400/401/403.
Log each retry with attempt number.
If all retries exhausted, return 502 with { error: 'AI service temporarily unavailable' }.
Unit test: mock OpenAI 429×2 then 200 → verify 200 response.
Unit test: mock OpenAI 500×3 → verify 502 response.

Acceptance: npm test passes. Wrapper is reusable if shared with contractpilot.

Generated by CEO Planner (priority: 3)

The /api/generate route calls OpenAI but has no retry logic — a single 429 or 500 fails the generation request. Implementation: 1. Wrap the OpenAI API call in a retry helper: max 3 attempts, backoff 1s → 2s → 4s. 2. Only retry on 429 and 500/502/503. Do NOT retry on 400/401/403. 3. Log each retry with attempt number. 4. If all retries exhausted, return 502 with `{ error: 'AI service temporarily unavailable' }`. 5. Unit test: mock OpenAI 429×2 then 200 → verify 200 response. 6. Unit test: mock OpenAI 500×3 → verify 502 response. Acceptance: `npm test` passes. Wrapper is reusable if shared with contractpilot. --- *Generated by CEO Planner (priority: 3)*