MOSAICO IP Agent

This agent uses an LLM to parse natural language queries about software libraries, extracts package details, and fetches real-time licensing information from the Eclipse Foundation and ClearlyDefined APIs.

Features

• Provider-Agnostic Extraction: Leverages LiteLLM to accurately extract package names, types (npm, pypi, maven), and versions. While optimized for Ollama, it supports 100+ LLM providers (OpenAI, Anthropic, Azure, etc.) via simple configuration changes.
• License Verification: Automatically queries the Eclipse Foundation and ClearlyDefined to find the declared license for the extracted package.
• A2A SDK Integration: Fully compatible with the MOSAICO A2A architecture, handling task states and artifact generation.
• Observability: Built-in tracing and observability using Langfuse.

Prerequisites

• Python: 3.10 or higher
• Package Manager: uv (recommended)
• LLM Backend:
  • Local: Ollama (Default)
  • Remote: Any LiteLLM-supported provider (OpenAI, Claude, etc.)

Installation

From PyPI

pip install mosaico-ip-agent

From GHCR (Docker)

docker pull ghcr.io/gomes89/mosaico-ip-agent:latest

From Source (Development Only)

git clone https://github.com/gomes89/mosaico-ip-agent.git
cd mosaico-ip-agent
uv sync

# Windows:
.venv\Scripts\activate 
# Linux/macOS:
source .venv/bin/activate

Configuration

The agent is configured using environment variables. You can set these in your terminal or use a .env file if you install a package like python-dotenv later.

Running the Agent

Because the project is installed as a package with defined entry points, you have two simple ways to start the agent:

Option 1: Using the CLI command

mosaico-ip-agent

Option 2: Running the module directly

python -m mosaico_ip_agent

You can verify the agent is running by querying the health check endpoint:

# Linux / macOS Bash:
curl -f http://localhost:9000/health

# Windows PowerShell:
Invoke-RestMethod -Uri "http://localhost:9000/health"

License

This program and the accompanying materials are made available under the terms of the MIT License, which is available at https://opensource.org/license/mit.