Ml Api Endpoint

What it does

This skill enables AI agents to design and deploy machine learning API endpoints, specifically using FastAPI for implementation. It focuses on creating robust APIs with features like stateless request handling, consistent response formatting, versioning strategies, and rigorous input validation before performing inference. The skill provides example code for both single prediction requests and batch predictions.

When to use it

• When an agent needs to expose a machine learning model's functionality as an API endpoint.
• For integrating machine learning models into automated workflows within an AI agent.
• To create scalable and maintainable ML services that can handle multiple requests.
• When batch prediction capabilities are required for processing larger datasets.

Key capabilities

• API Design Principles: Adherence to stateless design, consistent response formats, versioning strategies, and input validation.
• FastAPI Implementation: Utilizes FastAPI for building the API endpoints.
• Single Prediction Endpoint: Provides an endpoint (/predict) for individual prediction requests with input validation using Pydantic models.
• Batch Prediction Endpoint: Offers a batch prediction endpoint (/predict/batch) to process multiple inputs simultaneously, limited to a maximum of 100 instances.
• Model Loading: Automatically loads a pre-trained machine learning model (assumed to be in model.pkl) at startup.

Example prompts

• "Create an API endpoint for my machine learning model."
• "Design a batch prediction API that accepts up to 100 data points."
• "Implement input validation to ensure the features list has exactly 10 elements."

Tips & gotchas

• Requires familiarity with Python and FastAPI.
• The skill assumes a pre-trained machine learning model is available in model.pkl. You'll need to train and save your model accordingly.
• Batch prediction requests are limited to a maximum of 100 instances due to the validate_batch_size validator.