How srinikraman/rag-llama_index-pinecone works
This is a compact command-line RAG application. On startup, main.py loads PDFs, configures OpenAI models, ensures a dimension-compatible Pinecone index, and builds a vector index. It then accepts questions, queries the index with hybrid top-6 retrieval and a context-restricted prompt, and prints answers. The README documents these capabilities; implementation details are concentrated in main.py, so the graph keeps stages distinct without implying separate modules. Edges are supported by the main.py excerpt or explicit README statements.
Main entry points: main.py Vector index builder · input PDF documents
Components 13
PDF ingestion
- PDF documents
input - PDF loader
main.py - Document objects
main.py
Index and storage
- Index management
main.py - Pinecone vector DB
- Vector index builder
main.py - OpenAI embeddings
main.py
Query and response
- Interactive query loop
main.py - Hybrid retriever
main.py - Context-only prompt
main.py - OpenAI answer model
main.py - Printed answer
main.py
Other components
- CLI user
Connections 13
- PDF documents → PDF loader: loads from
main.py - PDF loader → Document objects: produces
main.py - Document objects → Vector index builder: indexes
main.py - Index management → Pinecone vector DB: checks and configures
main.py - Vector index builder → Pinecone vector DB: stores vectors
main.py - Vector index builder → OpenAI embeddings: embeds documents
README.md - CLI user → Interactive query loop: enters question
main.py - Interactive query loop → Hybrid retriever: queries index
main.py - Hybrid retriever → Pinecone vector DB: searches vectors
README.md - Hybrid retriever → OpenAI answer model: provides context
README.md - Context-only prompt → OpenAI answer model: constrains answer
main.py - OpenAI answer model → Printed answer: generates
main.py - Interactive query loop → Printed answer: prints
main.py