wyoloservice-mcp 0.2.0

MCP Server for NeuralForgeAI and Train Service 2

🚀 NeuralForgeAI MCP Server (wyoloservice-mcp)

An advanced Model Context Protocol (MCP) server that empowers AI agents to seamlessly interact with the NeuralForgeAI YOLO training cluster and remote Samba datasets.

🌟 Features

• Stateful Credential Management: Securely stores and manages API and CIFS credentials (set_cluster_credentials) so agents only ask once.
• Remote Dataset Validation: Spins up ephemeral Docker containers (worker_executor) to mount remote CIFS shares and validate YOLO structures (check_dataset_path, validate_dataset_advanced).
• Comprehensive Cluster Monitoring: Fetches real-time telemetry from /health, /workers, and /tasks in a single parallelized call (get_cluster_status).
• Sweeper YAML Generation: Autogenerates NeuralForge "Sweeper v2" configuration files locally for user review, dynamically discovering dataset classes and metadata (generate_training_yaml).
• 1-Click Training Launch: Submits YAML configurations directly to the NeuralForge API and automatically injects the study_id back into the local YAML file for complete traceability (launch_training, get_study_details, cancel_study).

⚙️ Installation

Install the package directly via pip:

git clone https://github.com/wisrovi/wyoloservice2_mcp.git
cd wyoloservice2_mcp
pip install -e .

This will expose the global wyolo-mcp binary.

🔌 Connecting to your AI Assistant

Add the following to your AI Assistant's MCP configuration file (e.g. ~/.gemini/config/mcp.json or Claude Desktop config):

{
  "mcpServers": {
    "neuralforge-mcp": {
      "command": "wyolo-mcp"
    }
  }
}

🧠 Agentic Workflow (Built-in Intelligence)

This MCP server is designed to self-instruct the LLM. For instance:

• Project Enforcement: If you don't provide a project name, the agent knows it must ask you to conform to <project>_<dataset>.
• Auto-Discovery: When you ask "how is my training going?", the agent is instructed by the MCP docstrings to automatically scan your directory for .yaml files, extract the study_id, and fetch the status without you providing any IDs.

---

Author: William Rodriguez (Wisrovi)