RAG LLM using LangChainGO

This project implements a Retrieval-Augmented Generation (RAG) system using LangChainGo. The architecture consists of:

• MongoDB instance for storing user history
• PostgreSQL database with pgvector for document embeddings
• Ollama service for managing LLM interactions
  • nomic-embed-text for embedding
  • sailor2:8b as the LLM

Prerequisites

• Docker
• NVIDIA GPU and drivers for GPU-accelerated Ollama (optional).

Getting Started

1. Clone the Repository:

   git clone https://github.com/oniharnantyo/lanchaingo-rag-golang
   cd lanchaingo-rag-golang

2. Copy backend config:

   Run the Docker Compose setup:

   cd backend
   cp .env.example .env

3. Start Services:

   Run the Docker Compose setup:

   docker-compose up -d

4. Stopping Services:

   To stop the services, run:

   docker-compose down

API Documentation

Add Documents

Uploads a document to the backend for processing and embedding.

Request:

curl --request POST \
  --url http://localhost:8000/documents \
  --header 'content-type: multipart/form-data' \
  --form 'file=@/path/to/your/document.pdf'

Query

Sends a query to the backend and retrieves relevant results from the indexed documents.

Request:

curl --request POST \
  --url http://localhost:8000/queries \
  --header 'content-type: application/json' \
  --data '{
  "query": "your question here"
}'

License

This project is licensed under the MIT License. See the LICENSE file for details.