Spaces:
Runtime error
Runtime error
Commit History
-
fix(docker): single-stage build with pinned versions — eliminates timeout
EymenProblems with old Dockerfile: - Multi-stage build: pip ran twice (builder + runtime copy) - No system deps: faiss-cpu needs libgomp1 at runtime → segfault - 2 workers on 512MB free tier RAM → OOM kills - Version ranges forced pip resolver to run longer Fixes: - Single stage: pip runs once, no copy between stages - libgomp1 installed: faiss works without segfault - Pinned exact versions: pip resolves in seconds not minutes - 1 worker: stays within free tier 512MB RAM limit - --timeout-keep-alive 75: prevents Render proxy from dropping SSE7d3d0fc
-
fix(critical): correct Claude model name + remove Streamlit import
EymenBREAKING BUG: claude-sonnet-4-6 is NOT a valid model name. This caused every AI request to hang for minutes then fail. Fix: - claude-sonnet-4-6 → claude-sonnet-4-5 (valid model) - max_tokens: 2048 → 1024 (2x faster responses) - yfinance_timeout: 15.0 → 10.0 (fail faster on bad tickers) - Removed st.secrets import from settings.py (crashed FastAPI startup) - Clean env-only key resolution: os.getenv → pydantic SecretStr8adb971
-
config: update URLs with live Render backend
Eymencd816d5 -
feat(deploy): add Render blueprint, Vercel config, Edge SSE proxy
Eymen- render.yaml: one-click Render deployment blueprint with all env vars - vercel.json: correct root directory and Edge function config - app/api/chat/stream/route.ts: Edge Runtime SSE proxy (no timeout) - next.config.ts: API rewrites + SSE cache headers - DEPLOY.md: 3-step deployment guide - lib/api.ts: chat stream routes through /api/chat/stream Edge proxyb9cb500