docs/architecture.md

KnowFlow AI Architecture

High-Level Flow

User uploads documents
-> Document loader extracts text
-> Text cleaner normalizes content
-> Chunker splits text into retrieval units
-> Embedding model converts chunks into vectors
-> ChromaDB stores vectors locally
-> Retriever finds relevant chunks
-> Prompt builder creates RAG prompt
-> Cloud LLM generates grounded answer
-> Streamlit displays answer and sources

Main Components

• Streamlit UI
• Document ingestion layer
• Chunking layer
• Embedding layer
• Vector database layer
• Retrieval layer
• Cloud LLM client
• RAG orchestration pipeline
• Logging and observability layer