Federal Registry RAG Agent

An AI-powered Retrieval-Augmented Generation (RAG) system that provides intelligent access to US Federal Registry documents including executive orders, regulations, notices, and other government publications.

Features

• Real-time Data Pipeline: Automatically fetches and processes federal documents from the official Federal Register API
• Intelligent Search: AI agent with tool-calling capabilities to query and analyze federal documents
• Interactive Chat Interface: Natural language interface to ask questions about federal documents
• SQLite Database: Efficient local storage optimized for Hugging Face Spaces

Quick Start

1. Initialize System: Click "Initialize System" to set up the database
2. Run Data Pipeline: Click "Run Data Pipeline" to download recent federal documents
3. Start Chatting: Ask questions about federal documents, executive orders, regulations, etc.

Example Queries

• "What are the recent executive orders from the last 7 days?"
• "Find documents about artificial intelligence regulations"
• "Show me documents from the Department of Defense"
• "Search for climate change related documents"
• "What new regulations were published this month?"

Technical Architecture

• Agent System: OpenAI-compatible LLM with function calling capabilities
• Data Pipeline: Asynchronous processing of Federal Register API data
• Database: SQLite for document storage and retrieval

Configuration

The system requires an external LLM API key (OpenAI, Groq, etc.). Set your API credentials in the Hugging Face Spaces secrets:

• LLM_API_KEY: Your LLM provider API key
• LLM_BASE_URL: API endpoint URL (default: OpenAI)
• LLM_MODEL: Model name (default: gpt-3.5-turbo)

Data Source

Documents are sourced from the official Federal Register API, ensuring access to authoritative and up-to-date government publications.

Limitations

• Free tier limitations apply to external LLM API usage
• Data pipeline processes recent documents (last 7 days by default)
• Storage limited to SQLite database on HF Spaces

Development

This system demonstrates:

• Asynchronous Python programming
• RAG system architecture
• LLM agent design with tool calling
• API integration and data processing
• Real-time web interfaces

Example Flow: Behind the Scenes of Federal RAG Agent

1. Data Ingestion: Integrated the official Federal Register API to stream and parse new documents automatically.
2. Vectorization: Each document is chunked and converted into embeddings using an LLM-based encoder.
3. Retrieva: Embeddings are stored in a SQLite vector database (now migrating to Pinecone for scalability). At query time, the system performs a semantic search to fetch the most relevant passages.
4. Answer Generation: Retrieved passages are combined with the user’s question and passed to an LLM (Meta-Llama-3.2-3B-Instruct) with strict grounding instructions to ensure accuracy.

Live Demo

The Federal RAG Agentic Chatbot is live and accessible on Hugging Face Spaces

Try it out directly in your browser to interact with federal documents in real time. Perfect for learning modern AI application development patterns and government data integration.