laya
laya is a Python library that makes fast typed decisions such as choices, scores and yes/no answers over text in 100+ languages for developers building agents.
What it is good for
- Sorting incoming messages to the right team automatically
- Getting a yes/no or score decision from any text quickly
- Routing requests to the right model based on language and length
- Calibrating decision confidence using your own labeled examples
At a glance
- Who it is for
- Setup difficulty
- Developer ยท Requires programming
Level 5 of 5. You need to write code to use it.
- Cost
- Free and open source
- Type
- Framework or SDK
- Tasks
- Works with
- LangChain, LangGraph, LlamaIndex, CrewAI, Hugging Face, ONNX Runtime
How to get started with laya
- Install Python 3.10 or newer.
- Run
python -m pip install layain a terminal. - Import laya in your Python code and create an agent with
laya.Agent(). - Pass your text and a list of questions to the agent's predict call to get typed answers.
From the project's own documentation. If a step does not match what you see, the official pages have the latest instructions.
Before you try it
- Requires Python 3.10 or newer and you need to write code to use it.
- Optional features such as vision, audio, server and MCP need extra install packages.
- Fitting on your own labels needs at least about 35 labeled rows for the example questions.
- Calibration with very few rows gives wide, less certain prediction sets.
Momentum
- Rank this week
- #44 in builder tools
- Best rank
- #35 week 45, 2026
- Stars gained last week
- +322
- GitHub stars
- 35,586
GitHub numbers as of Nov 16, 2026. Updated weekly.
Ranking history
Official links
Questions
- Is laya free?
- Yes. It is free and open source (Apache-2.0 license).
- Do I need to know how to code to use laya?
- Yes. You need to write code to use it.
- What does laya work with?
- According to its own documentation, it works with LangChain, LangGraph, LlamaIndex, CrewAI, Hugging Face, and ONNX Runtime.
- How do I get started with laya?
- 1. Install Python 3.10 or newer. 2. Run python -m pip install laya in a terminal. 3. Import laya in your Python code and create an agent with laya.Agent(). 4. Pass your text and a list of questions to the agent's predict call to get typed answers.
- What should I know before trying laya?
- Requires Python 3.10 or newer and you need to write code to use it. Optional features such as vision, audio, server and MCP need extra install packages. Fitting on your own labels needs at least about 35 labeled rows for the example questions. Calibration with very few rows gives wide, less certain prediction sets.
- What is laya good for?
- Sorting incoming messages to the right team automatically, getting a yes/no or score decision from any text quickly, routing requests to the right model based on language and length, and calibrating decision confidence using your own labeled examples.
Similar agents to laya
- AnyJev
Turns any open language model into a typed decision maker rather than using laya's own non-autoregressive checkpoints.
- Gateway
Routes requests across many language models through one API, while laya routes between its own decision models.
- skills
A skill that teaches coding agents to build integrations with a decision API, not a library for making decisions.
- LangChain
A broad framework for building agents and LLM apps, while laya focuses on fast typed decisions.
- Haystack
A general framework for production LLM apps and agent workflows, not a single-pass decision engine.
- LangBot
LangBot is a self-hosted platform that connects AI models and agent workflows to chat apps like Slack, Discord, Telegram, and WeChat for teams and businesses.