Qdrant Timeout Fix Report — 2026-03-31

PRs: #335 (timeout 15→3 + health re-probe), #336 (timeout 3→1 + deploy warm-up) Status: DEPLOYED + VERIFIED

Changes

Fix	Before	After	File
Qdrant timeout	15s	1s	vector_store.py:137
Health re-probe	Startup only (stale)	Every 30s cache TTL	server.py (root endpoint)
Deploy warm-up	sleep 45s, check 1x	Loop 10s, max 120s	deploy-vps.yml

Verification Results (Post-Deploy)

Health Check 5/5

Test	Status	Qdrant	Qdrant Latency	PG Latency	Docs
1-5	healthy	ok	9.2ms	0.6ms	798

Search Test 5/5

Test	Hits	Latency
1	5	4,296ms (cold)
2	5	397ms
3	5	478ms
4	5	410ms
5	5	397ms

Before vs After

Metric	Before	After	Improvement
Health status	degraded (stale)	healthy	Fixed
Qdrant latency (health)	753.4ms (cached)	9.2ms (live)	82x
Search latency	6,000-16,000ms	397-478ms	15-33x
Worst case retry	3x15s = 45s	3x1s = 3s	15x
Deploy success	FAIL (45s timeout)	SUCCESS (loop 120s)	Fixed

CI Status

• PR #335: All checks pass, merged
• PR #336: 20/20 checks pass, merged
• Deploy: SUCCESS (warm-up loop worked)

Conclusion

All 3 fixes deployed and verified. GPT Actions should no longer timeout (worst case 3s retry vs 30s GPT timeout). Health endpoint now shows live Qdrant status instead of stale startup value.