DOCKTOR

DOCKTOR and MCPX

DOCKTOR is a powerful new module in DHTI that allows you to "plug in" traditional machine learning models and other tools as if they were apps. It uses a technology called MCPX (Model Context Protocol Gateway) to connect everything together.

What is it?

Imagine you have a specialized AI model that can detect skin cancer from an image, or a tool that calculates risk scores. DOCKTOR lets you package these tools into Docker containers and instantly make them available to your DHTI agents.

Why use it?

• Modularity: Keep your heavy ML models separate from your core application.
• Flexibility: Use any language or framework (Python, PyTorch, TensorFlow) inside the container.
• Agent-Ready: Your AI agents can "see" these tools and decide when to use them.

How it works

1. Containerize: Package your model as an MCP server in a Docker container (we call this a "docktor" container).
2. Connect: The MCPX gateway discovers these containers.
3. Use: Your main DHTI agent (the "elixir") can now call functions in the container, like analyze_skin_lesion(image).

CLI Commands

Manage your inference pipelines using dhti-cli docktor.

npx dhti-cli docktor [OP] [NAME] [FLAGS]

Operations

Install

Install a new tool from a Docker image.

npx dhti-cli docktor install skin-cancer-model --image my-registry/skin-model:latest

• --image: Docker image name.
• --model-path: Path to map model files (optional).

Remove

Remove an installed tool and stop its container.

npx dhti-cli docktor remove skin-cancer-model

List

List all active docktor tools.

npx dhti-cli docktor list

Restart

Restart a specific tool container.

npx dhti-cli docktor restart skin-cancer-model

Configuration

• By default, it connects to the dhti-mcpx-1 container. Use -c to specify a different container name.
• Environment variables can be passed with -e: dhti-cli docktor install model -e "API_KEY=123".

This feature brings the power of specialized clinical AI models into your GenAI workflow!