Hijaling is a Claude Code skill that combines Hy, JAX, and Outlines to generate structured, executable Lisp-style code from natural language.

Can I use this for Machine Learning?

Yes, it specifically integrates with JAX, allowing you to generate and JIT-compile numerical functions and neural network components instantly.

How does it ensure code validity?

It uses the Outlines library and Pydantic schemas to constrain the LLM's output to valid Hy s-expression syntax, preventing syntax errors.

What are the core dependencies?

Hijaling requires Hy, JAX, Outlines, Pydantic, and access to an LLM provider such as Anthropic or OpenAI.

Does it work with other shells?

While it features specific NuShell integration for structured pipelines, the generated Hy code can be executed in any environment with Python and Hy installed.

Hijaling Structured Hy & JAX Generator

Name: Hijaling Structured Hy & JAX Generator
Author: plurigrid

byplurigrid

•

데이터 과학 및 ML

Generates syntactically guaranteed Hy s-expressions and JAX-compatible functions using constrained LLM decoding.

Hijaling bridges the gap between natural language prompts and executable, high-performance code by combining the Hy Lisp dialect with JAX and Outlines. It utilizes Pydantic-based schemas to force LLMs to generate valid s-expressions, which are then converted into Python ASTs and JIT-compiled via JAX. This ensures that generated mathematical models, neural network layers, or algorithmic logic are not only syntactically correct but also optimized for performance, making it an essential tool for automated scientific computing and symbolic AI workflows.

주요 기능

01Direct JAX integration for JIT-compiled numerical functions

02Extensible module schema for complex function definitions

03NuShell support for structured data pipelines and CLI execution

04Seamless conversion from natural language to valid Hy (Lisp) code

05Constrained s-expression generation via Outlines and Pydantic schemas

068 GitHub stars

사용 사례

01Building self-operating symbolic AI systems that generate and execute their own logic

02Creating robust CLI tools that transform prompts into structured, executable Lisp code

03Automating the creation of JIT-optimized neural network layers from text descriptions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add plurigrid/asi hijaling

For use in Claude.ai and ChatGPT

Download Skill