RAG Project

A small Retrieval-Augmented Generation (RAG) project using LangChain, ChromaDB, and OpenRouter-compatible OpenAI models.

Project structure

• answer_generation.py - Builds a query from a user prompt, searches the persisted Chroma vector store, and generates a response with ChatOpenAI.
• ingetion_pipline.py - Loads .txt documents from docs/, splits them into chunks, creates embeddings, and persists a ChromaDB vector store.
• retrival_pipeline.py - Loads the persisted vector store and performs semantic retrieval from the collection.
• docs/ - Contains source text documents used to create embeddings and answer questions.
• db/chroma_db/ - Persisted Chroma vector store directory.
• .env - Environment variables for API keys (not committed).

Usage

1. Build or refresh the vector store

Run the ingestion pipeline to load documents, split them into chunks, and persist embeddings:

python ingetion_pipline.py

2. Retrieve documents

Use the retrieval pipeline to load the persisted Chroma store and perform semantic search:

python retrival_pipeline.py

3. Generate an answer

Run the main answer-generation script:

python answer_generation.py

This script reads relevant documents from the persisted vector store, combines them into a prompt, and queries the chat model to create an answer.