How sumith2004/productive-rag works
Productive RAG is a Streamlit document-question-answering app. Users create or select a persistent chatbot, upload a PDF or DOCX for indexing, then ask questions answered from retrieved document content. The documented ingestion path partitions mixed content, chunks it, summarizes mixed-content chunks, embeds them, and stores vectors; querying retrieves relevant chunks and generates an answer. The excerpts show the app calling the shared pipeline and session-file persistence. Pipeline stage links and capabilities not directly visible in the excerpts are based on the README; notebooks are omitted as alternate workflow interfaces.
Main entry points: app.py Streamlit app · rag_pipeline.py Mixed-content summary
Components 14
App experience
- Streamlit app
app.py - Session files
app.py
Document ingestion
- Ingestion pipeline
rag_pipeline.py - Document partition
rag_pipeline.py - Title chunking
rag_pipeline.py - Mixed-content summary
rag_pipeline.py - Text embeddings
rag_pipeline.py - Chroma vectors
rag_pipeline.py
Question answering
- Query pipeline
rag_pipeline.py - Chunk retrieval
rag_pipeline.py - Answer generation
rag_pipeline.py
Other components
- User
- Groq LLM
- HuggingFace model
Connections 15
- User → Streamlit app: uploads and asks
README.md - Streamlit app → Ingestion pipeline: starts ingestion
app.py - Ingestion pipeline → Document partition: partitions
README.md - Document partition → Title chunking: produces elements
README.md - Title chunking → Mixed-content summary: forms chunks
README.md - Mixed-content summary → Text embeddings: prepares documents
README.md - Text embeddings → HuggingFace model: embeds with
README.md - Text embeddings → Chroma vectors: stores vectors
README.md - Streamlit app → Session files: persists sessions
README.md - Streamlit app → Query pipeline: requests retrieval and answer
app.py - Query pipeline → Chunk retrieval: retrieves chunks
README.md - Chunk retrieval → Chroma vectors: searches vectors
README.md - Query pipeline → Answer generation: generates answer
README.md - Answer generation → Groq LLM: invokes model
README.md - Mixed-content summary → Groq LLM: summarizes mixed content
README.md