Smoke test for the AI streaming pipeline (ASK-213)
POST/ai/ping
Internal smoke endpoint that exercises the OpenAI client +
SSE plumbing end-to-end. Streams a short response (default
"pong") from GPT-4.1-nano as a series of text/event-stream
events: delta, usage, done. Used by curl -N and the
backend health checks to verify wiring; not part of any
product feature. Kept under /ai/* so it shares the routing
bypass (no 60s timeout) that real AI endpoints will need.
Request
Responses
- 200
- 400
- 401
- 500
SSE event stream. Event types: delta (text chunk),
usage (final token + cache counts), done (terminator),
error (mid-stream failure). Each event's data: payload
is JSON.
Bad request
Unauthorized
Internal server error