Digital You - RAG Evolution

Earlier we utilized "Context Stuffing" to create the Digital You. This project evolves that concept into a Retrieval Augmented Generation (RAG) system for better scalability and efficiency.

Architectural Evolution

1. Ingestion: Reads documents from /data.
2. Chunking: Splits text into manageable segments.
3. Embedding: Converts text into numerical vectors.
4. Retrieval: Fetches only relevant segments to answer user queries.

Setup

1. Sync Environment: uv sync in the root directory.
2. Data: Place .txt, .md, or .pdf files in the /data directory.
3. Tasks: Complete the TODOs in app.py to build the RAG pipeline.

Your Mission

1. Data Infrastructure: Build a vector store in chroma_db/ using your bio data.
2. Contextualization: Teach the AI how to re-write questions based on history so it never "forgets" the subject of conversation.
3. Persona Engineering: You must write your own qa_prompt. This is where you define your "Digital Twin" identity and force the AI to only use the retrieved context.

Enhancements

• Use SemanticChunker for smarter context splitting.
• Customize the Gradio UI using gr.Blocks.