RAG Client

Lightweight RAG (Retrieval-Augmented Generation) system that logs performance metrics to a remote server.

What This Does

• Embeds queries using sentence-transformers
• Searches documents using Qdrant vector database
• Generates answers using OpenAI
• Logs all performance metrics to a remote logging server

Setup

1. Clone the repo:

git clone https://github.com/al-gent/rag-client.git
cd rag-client

2. Create .env file:

cp .env.example .env
# Edit .env with your values

Required variables:

• RAG_HARDWARE_ID - Identifier for your hardware (e.g., laptop-mac, server-gpu)
• LLM_HARDWARE_ID - Where LLM runs (e.g., openai-api, local-gpu)
• MODEL_NAME - Which model to use (e.g., gpt-4o-mini)
• LOG_SERVER_URL - Remote logging server (e.g., https://rag-api.adamlgent.com)
• OPENAI_API_KEY - Your OpenAI API key

3. Start the system:

docker-compose up -d

4. Load documents:

# Put PDF or TXT files in ./data directory
curl -X POST http://localhost:8000/load-documents

5. Query:

curl -X POST http://localhost:8000/query \
  -H "Content-Type: application/json" \
  -d '{"question": "Your question here"}'

What Gets Logged

Every query logs:

• Embedding time
• Vector search time
• LLM generation time
• Total time
• Cost
• Success/failure

All metrics are sent to the remote logging server for comparison across different hardware setups.

Endpoints

• POST /query - Ask a question
• POST /upload - Upload a document
• POST /load-documents - Load all documents from ./data
• GET /health - Health check
• GET /documents - List indexed documents