RAG Agent with ChromaDB and Web Search

This project implements a Retrieval-Augmented Generation (RAG) agent that uses ChromaDB as the vector store and performs web search to ingest documents. The agent answers user questions by retrieving relevant passages from the stored documents and generating responses with OpenAI’s GPT models.

Features

ChromaDB vector store (no Qdrant usage)
Web content ingestion via HTTP fetch
OpenAI embeddings for vector representation
GPT-4o-mini for answer generation
Simple CLI usage

Prerequisites

Node.js 20+ (ESM support)
Docker (optional, for running ChromaDB locally)
OpenAI API key

Setup

Clone the repository

git clone https://git.brojs.ru/kuzakhmetovartur/ekzamen-rag-agent-s-chromadb-i-veb-poisk.git
cd ekzamen-rag-agent-s-chromadb-i-veb-poisk

Install dependencies
```
npm install
```

Configure environment variables

Create a .env file in the project root:

CHROMA_HOST=localhost
CHROMA_PORT=8000
OPENAI_API_KEY=YOUR_OPENAI_API_KEY

Run ChromaDB

The simplest way is to use Docker:
```
docker run -d --name chromadb -p 8000:8000 chromadb/chroma
```
Or install ChromaDB locally following the official docs.
Run the agent
```
npm start
```
The script will ingest a sample document from GitHub and answer a question about the OpenAI Node.js library.

Project Structure

src/
├── agent.js        # RAG agent logic
├── index.js        # Entry point
├── vectorStore.js  # ChromaDB wrapper
└── webSearch.js    # Simple web fetch helper

Notes

The project does not use Qdrant. All references to Qdrant have been removed.
Only ChromaDB is used for vector storage.
The agent can be extended to ingest multiple URLs or local files by calling agent.ingestFromUrl(url).

License

MIT License

Feel free to contribute or open issues for enhancements.

2.0 KiB Raw Blame History Unescape Escape