Skip to content
AgentAgent framework

humansignal

Open-source Python framework for building autonomous agents that label and process data

humansignal is an open-source agent framework for developer tools. It is available on API, Self-hosted, Linux, macOS and Windows.

Pricing
Open source
Platforms
API, Self-hosted, Linux
GitHub stars
1.6K
Last commit
Language
Python
License
Apache-2.0

Similar agents

About humansignal

Adala is a Python library for building agents that take on data processing work, with most of its focus on labeling datasets. Instead of hand-writing prompts for every label, you give an agent a ground truth dataset as its environment, and it iteratively refines the skills it uses against that data.

The framework is organised around a few building blocks: agents, environments that hold the ground truth signal, datasets, memories, runtimes that wrap LLM providers, and skills such as classification, summarization, translation and question answering. A single skill can run on several runtimes, which allows a cheaper model to do the work while a stronger one acts as a teacher during learning.

It suits ML engineers, data scientists and researchers who want to run labeling or preprocessing jobs from a Python notebook or script. The README also points to OpenRouter as a way to use Claude or Gemini models through an OpenAI-compatible interface.

What it does

Autonomous skill learning
Agents iteratively refine their skills against a ground truth dataset over a set number of learning iterations.
Prebuilt skill types
Classification, summarization, translation, question answering and text generation skills ship with the library.
Teacher and student runtimes
A skill can run on one model while a separate teacher runtime guides learning.
Pluggable LLM runtimes
Runtimes wrap OpenAI-compatible chat APIs, and OpenRouter can route requests to other model providers.
Pandas dataset support
Training and test data are passed in as pandas DataFrames for batch labeling.

Good for

  • Labeling a text dataset into predefined sentiment or topic categories
  • Summarizing or translating batches of text inside a notebook workflow
  • Preprocessing and postprocessing large DataFrames with LLM-based skills
  • Teaching or experimenting with agent design and skill decomposition

Pricing

Open source

Open source under Apache-2.0; you pay your own LLM provider costs.

Prices change. Check humansignal.github.io before you buy.

humansignal FAQ

How do I install Adala?

Install it with pip install adala. The README also shows installing from GitHub for the latest updates and a developer install using poetry.

Which LLM providers does humansignal support?

The examples use OpenAI chat models. The README also shows using OpenRouter to reach Claude and other OpenAI-compatible models.

Where can I get community help?

The README links to a Discord channel for questions and discussion.

Is humansignal free?

Yes. humansignal is open source, so you can use and self-host it for free.

What are the best alternatives to humansignal?

The closest alternatives to humansignal are Web, devika, pydantic-ai, DemoGPT and ag2.

Where this comes from

  • Description: written from the project's README and the project's website.