fix(rag): add sse keepalive heartbeat to prevent nginx 60s timeout

[AndreLiar]

Root cause

The RAG /stream endpoint uses SSE. The embedding phase (multi-query expansion via CPU-only ollama) takes 3–5 minutes on the current prod infra. Nginx's default proxy_read_timeout is 60 seconds — it kills any SSE connection where no body bytes are written within that window.

Result: the SSE stream was dropped silently 60s after opening. The frontend stayed on "Retrieving context…" indefinitely, then reset to 0 messages. The backend actually completed the pipeline successfully (confirmed in logs: totalTime: 323330ms, messages saved to DB) but no client was connected anymore to receive the events.

Fix

Send an SSE comment (: keepalive\n\n) every 20 seconds from the moment the stream opens. SSE comments are valid per spec, ignored by browsers, and count as body bytes — so nginx resets its idle timer on each one, keeping the connection alive for the full pipeline duration.

const heartbeat = setInterval(() => {
  if (!closed && !res.writableEnded) res.write(': keepalive\n\n');
}, 20000);
// cleared in finally block

Test plan

• All 1449 backend unit + integration tests pass
• All 521 frontend tests pass
• E2E: send a question via the Ask AI UI → response appears within 5–6 minutes (not stuck at "Retrieving context…")

Notes

The underlying performance issue (CPU-only ollama at ~5 min/query) is a separate infra concern. This fix makes the feature functional while that is addressed.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 0% with 3 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
backend/controllers/ragController.js	0.00%	3 Missing ⚠️

📢 Thoughts on this report? Let us know!