How bhardwajtanuj/build-a-multimodal-rag-system-using-aws-bedrock-and-faiss works
This is a multimodal retrieval-augmented generation application with CLI and Streamlit entry points. Users ingest documents and images into a FAISS index, then submit text or image queries to retrieve relevant content and generate an answer with Bedrock. The README documents Nova/Titan embedding options, supported input formats, and Claude generation; source excerpts show the entry-point wiring, but not the internals of the src modules, so those links rely on documented architecture.
Main entry points: app.py Streamlit UI · main.py CLI
Components 10
User interfaces
- CLI
main.py - Streamlit UI
app.py
RAG pipeline
- Ingestion
- RAG query
Models and storage
- Bedrock embeddings
- FAISS index
- Configuration
Other components
- User
- AWS Bedrock
- Claude generation
Connections 18
- User → CLI: runs commands
README.md - User → Streamlit UI: uses interface
README.md - CLI → Ingestion: runs ingestion
main.py - Streamlit UI → Ingestion: runs ingestion
app.py - CLI → RAG query: runs queries
main.py - Streamlit UI → RAG query: submits queries
app.py - CLI → Bedrock embeddings: creates
main.py - CLI → FAISS index: loads index
main.py - Streamlit UI → Bedrock embeddings: creates
app.py - Streamlit UI → FAISS index: creates
app.py - Streamlit UI → Configuration: reads paths
app.py - Ingestion → Bedrock embeddings: embeds content
README.md - Ingestion → FAISS index: stores vectors
README.md - RAG query → Bedrock embeddings: embeds query
README.md - RAG query → FAISS index: retrieves matches
README.md - RAG query → Claude generation: generates answer
README.md - Bedrock embeddings → AWS Bedrock: requests embeddings
README.md - Claude generation → AWS Bedrock: uses model
README.md