🦑 Hyjinx

Functions and macros useful for modern living in the Hy REPL.

Compatible with Hy>=1.0

Hylights

• A sensible, lispy defmethod (polymorphic type dispatch)
• A Hy code beautifier and in-REPL syntax highlighting -- now moved to beautifhy
• Pretty tracebacks with correct syntax highlighting for hy or python -- now moved to beautifhy
• An inspect module that works like Python's, but also for Hy code
• Print/get/edit source code of a function, module etc.; e.g. (print-source hyjinx.lib.sieve)
• In-repl code analysis and discussion using LLMs (e.g. for writing docstrings, code review)
• In-repl syntax highlighting and latex rendering (for sixel-capable terminals) for LLM replies
• Numpy array pretty printing
• A zmq lazy pirate protocol
• A minimal ncurses class
• An async actor-model class

Modules

• hyjinx.lib: a smorgasbord of convenience functions.
• hyjinx.source: live code inspection.
• hyjinx.inspect: code inspection.
• hyjinx.doc: peruse hy documentation.
• hyjinx.actors: the async Actor class and defactor macro
• hyjinx.screen: a convenient ncurses wrapper.
• hyjinx.mat: numpy pretty-printing for humans. (requires numpy, jax optional)
• hyjinx.[zmq_client, zmq_server, crypto, wire]: lazy-pirate zmq RPC architecture. (requires zmq, ecdsa, zstandard, msgpack)
• hyjinx.llm: discuss code with a Large Language Model (AI). TabbyAPI, OpenAI-compatible and Claude are supported. (requires openai and anthropic packages.)

Install

$ pip install -U hyjinx

or, with optional dependencies,

$ pip install -U hyjinx[zmq]
$ pip install -U hyjinx[llm]

To install offline hy/hyrule documentation,

$ hy -m hyjinx.docs

For in-terminal sixel rendering of latex in LLM replies, make sure pdflatex, dvipng and img2sixel are installed, and that you're using a sixel-capable terminal.

Optional dependencies

You can install with the [zmq] option which also installs ecdsa, pyzmq and zstandard for the zmq server/client

You can install with the [llm] option which also installs openai and lets you discuss code objects with ChatGPT or a locally-served LLM (via TabbyAPI or similar).