LLM Agent Development Environment

This repository contains a Docker-based development environment for LLM agents using smolagents.

Setup

Make the startup script executable:

chmod +x scripts/start.sh

Set the necessary environment variables:

export OPENROUTER_API_KEY=your_api_key_here
export OPENAI_API_KEY=your_openai_api_key_here
export QDRANT_API_KEY=your_qdrant_api_key_here
export TELEMETRY_ENABLED=true

Running with Docker Compose

Build and start the container:

docker compose up --build

This will:

• Start the Phoenix telemetry server on http://localhost:6006
• Start the Gradio UI on http://localhost:7860

To shut it down, use CTRL+C and then:

docker compose down

Directory Structure

• /data: You should put your source data here.
• /agent_workspace: This is where the code produced by the agents will be stored.
• /scripts: Contains the python scripts to run the agents service.

Prompting the CatalystNeuro Agents

Some useful prompt templates can be found in the scripts/prompts/ directory. Adapt these to your use case.