Skip to content

zyx

ZYX Hero ZYX Hero

A fun "anti-framework" for doing useful things with agents and LLMs.

PyPI version Python version License

What is ZYX?

ZYX is a simplicity-first library wrapped on top of Pydantic AI and heavily inspired by Marvin. It aims to provide a simple, stdlib-like interface for working with language models, without loss of control as well as flexibility that more complex frameworks provide.

Introduction

ZYX stands on the shoulders of Pydantic AI to provide a set of functions and components that aim to complete the following goals:

  • Simplicity-First: Above all else, ZYX aims to be as simple as possible to use. The library is designed in a very semantically literal sense, with the hope that a 12 year old could pick up and run with it immediately.
  • Fast & Flexible: ZYX is designed to be as fast as possible in two senses, (1) performance wise, and (2) development time wise, providing a very flexible interface that allows rapid prototyping and iteration.
  • Type-Focused: ZYX provides a very type-focused interface, leveraging Pydantic's powerful validation capabilities to ensure that your data is always in the expected format.
  • Model Agnostic: Through Pydantic AI, ZYX is completely model agnostic and supports virtually any LLM provider.

Getting Started

The following guide is a quick way to get started on how to install ZYX and get up and running with a few of the core concepts.

Installation

You can install ZYX using your favorite Python package manager.

pip install zyx

uv add zyx

poetry add zyx

conda install zyx
Adding Model Providers

The core package for ZYX installs the minimum required dependencies from Pydantic AI using their pydantic-ai-slim package, along with the openai library for out of the box OpenAI support.

To add support for additional LLM providers, you can use one of the following methods to install the relevant dependencies:

To install all of pydantic-ai's supported providers, simply install the full package along with ZYX:

pip install zyx pydantic-ai

Or use the ai extra:

pip install 'zyx[ai]'

If you dont want to install the entire pydantic-ai package, you can easily add your desired providers individually, either through installing their libraries directly, or through a ZYX extra:

# install provider libraries directly
pip install zyx anthropic

Quickstart

Once you've installed ZYX, try running some of the following code examples to get a feel for the library!

Generating Content

The easiest way to use LLMs within ZYX is through the zyx.make semantic operation. Semantic operations are specialized functions that allow you to use LLMs to generate/edit content, parse data, and more!

Generating Content
import zyx


result = zyx.make(
    target=int, # (1)!
    context="What is 45+45?",
)


print(result.output)
"""
90
"""
  1. Targets are how to define the output of a semantic operation. In this case, we want to generate an integer.

Parsing a Website

Through the source, context and attachments parameters, ZYX provides a bunch of options on how you want to pass content to a model.

Parsing a Website
import zyx
from pydantic import BaseModel


class Information(BaseModel): # (1)!
    library_name : str
    library_description : str


def log_website_url(url : str) -> None:
    print(f"Website URL: {url}")


result = zyx.parse( # (2)!
    source=zyx.paste("https://zyx.hammad.app"), # (3)!
    target=Information,
    instructions="log the website URL before you parse.", # (4)!
    model="openai:gpt-4o-mini", # (5)!
    tools=[log_website_url] # (6)!
)


print(result.output.library_name)
print(result.output.library_description)
"""
Website URL: https://zyx.hammad.app
ZYX
A fun "anti-framework" for doing useful things with agents and LLMs.
"""

  1. First-class support for structured outputs in nearly every semantic operation within ZYX.

  2. zyx.parse is a semantic operation that allows you to parse a source object or content into a target.

  3. zyx.paste is a convenience function for creating a Snippet from a source object.
    A Snippet is a piece of content that can be used to represent multimodal or textual content easily.

  4. The context parameter is how prompts and messages are provided to the model. ZYX provides many options on how context can be provided!

  5. All models and providers supported by Pydantic AI are supported within ZYX.

  6. Thanks to Pydantic AI, all semantic operations inherently support tool usage.