A hyper-fast, fun, quality-of-life focused & genuinely useful LLM toolkit. Inspired by Marvin-AI. Built on LiteLLM, Instructor & Qdrant.
Project description
zyx
zyx 0.3.00 - the first 'release' of the library is out!
All CrewAI & other obstructions have been removed, and the library is back to being lightweight.
zyx
is a hyper-fast, fun, & ease-of-use focused Python library for using LLMs.
It was created on top of Instructor
and LiteLLM
, and focuses to provide an abstraction free framework.
The library uses methods such as lazy-loading to provide a single import for all its features. This library is not meant to be used as a production-ready solution, but rather as a tool to quickly & easily experiment with LLMs.
Some of the key features of zyx
include:
- Universal Completion Client : A singular function that handles all LiteLLM compatible models, Pydantic structured outputs, tool calling & execution, prompt optimization, streaming & vision support.
- A Large Collection of LLM Powered Functions : This library is inspired by
MarvinAI
, and it's quick LLM function style framework and has built upon it vastly. - Easy to Use Memory (Rag) : A
Qdrant
wrapper, built to support easy store creation, text/document/pydantic model data insertion, universal embedding provider support, LLM completions for RAG & more. - Multimodel Generations : Supports generations for images, audio & speech transcription.
- Functional / Easy Access Terminal Client : A terminal client built using
textual
to allow for easy access tozyx
features. - New Experimental Conversational Multi-Agent Framework : Built from the ground up using
Instructor
, the agentic framework provides a solution towards conversationally state managed agents, with task creation, custom tool use, artifact creation & more.
Getting Started
Installation
pip install zyx
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.