Which inference function types can I design with this skill?

The skill supports five specific types: MetaFn (metadata), TrigFn (trigger logic), PostFn (scoring/formatting), Src2InputFn (data transformation), and Input2SrcFn (inference entry mapping).

How does the testing feature work?

By using the '/haipipe-end test' command with a payload path, the skill runs an inference pass and provides performance profiling data to ensure the endpoint meets requirements.

Does it support cloud deployment?

Yes, haipipe-end specifically supports deployment to Databricks Unity Catalog and Model Serving, as well as local deployment for testing.

Why does the skill require a 'Bootstrap' step?

The bootstrap step ensures Claude reads the mandatory architecture and YAML configuration files (0-overview.md) to maintain context and accuracy throughout the deployment protocol.

What is the primary purpose of haipipe-end?

It is a specialized skill for the final stage of ML model deployment, handling the packaging of model instances into endpoints, testing their inference logic, and deploying them to production environments.

Haipipe Endpoint Deployment

Name: Haipipe Endpoint Deployment
Author: jluo41

byjluo41

0•

データサイエンスとML

Streamlines machine learning model endpoint deployment through unified packaging, inference function design, and multi-target deployment.

Haipipe-end is a specialized Claude Code skill designed for Stage 6 of the machine learning lifecycle: transitioning from ModelInstance_Sets to production-ready Endpoint_Sets. It provides a structured framework for designing the five essential inference function types—Meta, Trigger, Post-processing, Source-to-Input, and Input-to-Source. By enforcing a rigorous step-by-step protocol, the skill helps developers package assets, test inference with detailed profiling, and deploy seamlessly to Databricks Unity Catalog or local environments while maintaining architectural consistency.

主な機能

01Interactive status dashboard for monitoring EndpointStore and architecture overview.

02Specialized builders for MetaFn, TrigFn, PostFn, Src2InputFn, and Input2SrcFn.

030 GitHub stars

04Multi-target deployment support for Databricks Model Serving and local testing.

05Unified management of the Stage 6 Endpoint_Set deployment lifecycle.

06Integrated inference testing with automated profiling and payload validation.

ユースケース

01Designing and reviewing custom pre-processing and post-processing logic for AI pipelines.

02Profiling inference performance and validating response schemas before final deployment.

03Packaging trained model instances into production-ready endpoint sets for Databricks.

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add jluo41/tools haipipe-end

For use in Claude.ai and ChatGPT

Download Skill