How ferozewaris/multi-agent-intelligent-assistant works
Intelligent RAG System combines category-based PDF retrieval with structured-data analysis, exposed through a Streamlit chat UI and FastAPI. The documented ingestion path loads and splits categorized documents, embeds chunks, and stores them; queries are routed to RAG or DuckDB, then returned to the user. The graph includes the API, UI, indexing pipeline, model configuration, and MLflow tracking. Some UI routing and API indexing details are in omitted excerpts, so those links are marked unverified or omitted rather than asserted as observed calls.
Main entry points: rag_chain.py RAG Chain · duckdb_agent.py DuckDB Agent · embed_documents.py Document Indexer · streamlit_app.py Streamlit Chat
Components 16
User Interfaces
- Streamlit Chat
src2/streamlit_app.py - REST API
src2/api.py
Document Retrieval
- Document Indexer
src2/embed_documents.py - PDF Processing
src2/document_processor.py - Embedding Provider
src2/embeddings.py - Category Vector Store
src2/vector_store.py - RAG Chain
src2/rag_chain.py - Categorized PDFs
src2/data
Structured Data
- Agent Router
src2/agent_router.py - DuckDB Agent
src2/duckdb_agent.py - DuckDB
src2/duckdb_agent.py
Shared Services
- Runtime Configuration
src2/config.py - MLflow Tracking
Other components
- User
- Groq Chat Model
- Embedding Services
Connections 25
- User → Streamlit Chat: asks query
README.md - User → REST API: sends request
README.md - Streamlit Chat → Agent Router: selects agent
src2/streamlit_app.py - Streamlit Chat → RAG Chain: queries
src2/streamlit_app.py - Streamlit Chat → DuckDB Agent: handles data query
src2/streamlit_app.py - REST API → RAG Chain: queries
src2/api.py - Document Indexer → PDF Processing: loads and splits
src2/embed_documents.py - Categorized PDFs → PDF Processing: provides PDFs
src2/document_processor.py - Document Indexer → Embedding Provider: initializes
src2/embed_documents.py - Document Indexer → Category Vector Store: adds chunks
src2/embed_documents.py - RAG Chain → Category Vector Store: searches categories
src2/rag_chain.py - RAG Chain → Embedding Provider: uses embeddings
src2/rag_chain.py - Agent Router → Groq Chat Model: classifies query
src2/agent_router.py - DuckDB Agent → DuckDB: executes SQL
src2/duckdb_agent.py - DuckDB Agent → Groq Chat Model: generates SQL
src2/duckdb_agent.py - DuckDB Agent → Groq Chat Model: summarizes results
src2/duckdb_agent.py - RAG Chain → Groq Chat Model: generates answer
src2/rag_chain.py - Embedding Provider → Embedding Services: creates vectors
src2/embeddings.py - Agent Router → MLflow Tracking: logs routing
src2/agent_router.py - DuckDB Agent → MLflow Tracking: tracks queries
src2/duckdb_agent.py - Document Indexer → MLflow Tracking: tracks indexing
src2/embed_documents.py - RAG Chain → MLflow Tracking: tracks retrieval
- Runtime Configuration → Agent Router: configures
src2/agent_router.py - Runtime Configuration → Embedding Provider: configures
src2/embeddings.py - Runtime Configuration → Category Vector Store: configures
src2/vector_store.py