Spaces:
Sleeping
Sleeping
Commit History
-
Simplify deployment: drop Postgres/RAG, plain uvicorn on lean requirements
The previous Dockerfile installed PostgreSQL + pgvector via apt and had start.sh boot postgres before uvicorn. That path was failing on the Space (Runtime error: Scheduling failure / crash before serving), likely because the container can't run postgresql as a system service in the Space sandbox. This reverts to the working simple setup: python:3.11-slim, lean requirements (no jupyter/mlflow/optuna/postgres deps), and a freshly built Angular frontend.9567d15
-
Fix: pin python:3.11-slim-bookworm to avoid trixie/pgdg dependency conflict
=Apyhtml200af5920 -
Fix Docker build: use prebuilt postgresql-17-pgvector from official APT repo
=Apyhtml20869d567 -
Add RAG module: PostgreSQL + pgvector in same container
=Apyhtml20 +1: Claude Sonnet 4.6- PostgreSQL 16 + pgvector installed in Docker image - start.sh initializes DB and starts FastAPI on port 7860 - Full RAG pipeline: TF-IDF/SVD embeddings + pgvector search - NVIDIA NIM justification (requires NVIDIA_API_KEY secret) - Multi-model inference: logistic, lgbm, xgb - train.json (16MB) tracked via git-lfs - Angular SPA served from FastAPI Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>45fa780 -
Deploy Triple-R Checker: FastAPI + Angular frontend
=Apyhtml20 +1: Claude Sonnet 4.6- FastAPI serves Angular SPA on port 7860 - All ML models included (logistic, lgbm, xgb + vectorizer) - Angular build with relative API URL (/predict) - Docker Space configuration Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>9861b96