What’s New:

• Multi-hop Queries: The system now breaks down complex questions into simpler sub-queries and performs iterative retrieval across multiple “hops” for more comprehensive answers
• Negation Handling: Smart processing of queries with negation words like “not”, “without”, “except” to filter out unwanted results
• Vector Search: Uses OpenAI embeddings for semantic similarity search, making document retrieval much more accurate
• Seamless Integration: Works perfectly with our existing chunking strategies - just load your documents and start querying!

fixes: #34 /claim #34

Key Features:

Handles complex multi-step questions intelligently
Processes negation queries effectively
Integrates with existing Unsiloed chunking
Simple API - just call rag.query() with your documents loaded

Files Changed:

• Added Unsiloed/services/rag.py - Main RAG system implementation

• Updated Unsiloed/__init__.py - Exposed RAG functionality

• Created example_rag.py - Usage examples

• Updated README.md - Added documentation

• This makes Unsiloed much more powerful for users who need to ask sophisticated questions across their document collections!