Local-RAG

Local-RAG is a monorepo for a local Retrieval-Augmented Generation (RAG) system, designed to enable knowledge-intensive chat and document search workflows using modern LLMs and vector databases.

Structure

• apps/backend: A NestJS backend providing APIs for chat, document ingestion, retrieval, and thread management. Integrates with ChromaDB for vector search and supports streaming responses from local LLMs via Ollama.
• apps/frontend: An Angular frontend for interactive chat, thread management, and document search. Features markdown rendering, source attribution, and a responsive UI.
• packages/: Shared libraries and utilities for backend/frontend integration.

Features

• Document ingestion and chunking (PDF, text, etc.)
• Embedding and retrieval using ChromaDB and LangChain
• Local LLM integration via Ollama (configurable models)
• Multi-threaded chat with source attribution
• Streaming responses for improved UX
• Configurable chunk size, overlap, and accuracy threshold
• SQLite-based metadata storage

Getting Started

1. Install dependencies

   pnpm install

2. Configure environment
   Copy .env.example to .env and adjust settings as needed (ChromaDB, Ollama, etc).

3. Start backend

   pnpm dev:backend

4. Start frontend

   pnpm dev:frontend

Configuration

See .env.example for all environment variables, including database paths, model names, and service URLs.

Technologies

• NestJS, Angular, LangChain, ChromaDB, Ollama, SQLite

License

UNLICENSED (see package.json)