title | emoji | colorFrom | colorTo | sdk | sdk_version | app_file | pinned |
---|---|---|---|---|---|---|---|
HH Agent Tails |
🦔 |
indigo |
purple |
gradio |
4.14.0 |
hh_agent_tails/app.py |
false |
HH Agent Tails is an AI-powered documentation assistant for the Hedgehog Open Network Fabric project. It uses advanced RAG (Retrieval-Augmented Generation) techniques to provide accurate answers about Hedgehog's features, capabilities, and supported devices.
- Smart Document Search: Uses vector similarity search with Supabase pgvector to find relevant documentation
- Context-Aware Responses: Provides answers based on official Hedgehog documentation with source citations
- User-Friendly Interface: Simple chat interface built with Gradio
- Production Ready: Deployed on Hugging Face Spaces with read-only mode for safe public access
- Vector Store: Supabase with pgvector extension for efficient similarity search
- Embeddings: OpenAI embeddings for document vectorization
- LLM: GPT-3.5-turbo for natural language understanding and response generation
- Document Processing: Handles JSON documentation with 1500-token chunks and 300-token overlap
- Backend: FastAPI for robust API endpoints
- Frontend: Gradio for the chat interface
Simply type your questions about Hedgehog in the chat interface. The assistant will search through the documentation and provide relevant answers with citations to source documents.
Example questions:
- "What leaf switches does Hedgehog support?"
- "Does Hedgehog support the Edgecore DCS203?"
- "How does Hedgehog handle multi-tenancy?"
To run locally in development mode:
# Set up environment
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
# Run in development mode
ADMIN_MODE=true APP_ENV=development .venv/bin/python -m uvicorn hh_agent_tails.app:app --reload
The app automatically deploys to Hugging Face Spaces when changes are pushed to the main branch. In production, the app runs in read-only mode for security.
This project is part of the Hedgehog Open Network Fabric ecosystem and follows its licensing terms.