Apppublic

saadi786/ros2-rag-backend

sourceHugging Faceupdated 8mo agoView on Hugging Face

0likes

App README

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

Agentic RAG Backend

This project contains the backend for a RAG-powered conversational agent for the ROS 2 Textbook.

To run the server, use the canonical run script:

bash

./scripts/run_server.sh

Important: The GEMINI_API_KEY environment variable must be set for the application's LLM functionality to work correctly.

This script starts the FastAPI application with the recommended stable settings.

—Uvicorn Workers: The number of uvicorn workers is intentionally set to 1. Increasing the number of workers has been observed to cause silent runtime crashes in some environments. This is likely due to how certain libraries (e.g., qdrant-client, sentence-transformers) handle multiprocessing.
—WSL (Windows Subsystem for Linux): When running in WSL or other non-native Linux environments, it is critical to use the hardened flags provided in the run_server.sh script (--workers 1 --loop asyncio --http h11) to ensure stability. Using default uvicorn settings in these environments may lead to unexpected silent crashes.

>>>>>> c55823b (feat: Implement Gemini API integration improvements and production hardening)