Repolode
Submit
WecoAI

WecoAI/aideml

An open-source LLM agent that drafts, runs, debugs and iterates on machine learning code via tree search until a user-defined metric is optimized.

View on GitHubWebsite
Stars
1,614
Forks
237
Open issues
3
Last commit
22 days ago

What is it?

AIDE ML is the open-source reference implementation of the AIDE algorithm, a tree-search agent that treats each Python script as a node in a solution tree. Given a dataset directory, a goal described in plain language, and an evaluation metric, it generates code, executes it, and uses the metric feedback to decide which candidates to extend or fix.

The project ships as a Python package installable with pip, along with a command-line interface, an HTML viewer for the solution tree, and a Streamlit web UI. It works with OpenAI, Anthropic, Gemini, OpenRouter, or any local model that exposes an OpenAI-compatible API, and configuration covers the coding model, feedback model, step count and number of drafts. A Python API and Docker image are also provided.

The codebase is intended for researchers who want to replicate the AIDE paper or test new search heuristics, evaluators, or LLM backends, and for practitioners who want a starting ML pipeline built from a dataset.

Key features

  • 01Describes the task in plain language, with a dataset directory, goal and evaluation metric.
  • 02Searches a tree of generated Python scripts, extending or debugging candidates based on metric feedback.
  • 03Provides a CLI, a Python API, and a Docker image.
  • 04Includes an HTML viewer for the solution tree and a Streamlit web UI.
  • 05Works with OpenAI, Anthropic, Gemini, OpenRouter, and local OpenAI-compatible models.

Use cases

  • Prototyping a baseline ML model on a tabular dataset from a plain-language goal.
  • Replicating or extending the AIDE tree-search experiments in research.
  • Testing alternative search heuristics, evaluators, or LLM backends for ML agents.
  • Running a fully local pipeline against a model served by Ollama.

Quick start

  1. 1

    Install the package from PyPI

    pip install -U aideml
  2. 2

    Set your LLM API key

    export OPENAI_API_KEY=<your-key>
  3. 3

    Run an optimisation on the example task

    aide data_dir="example_tasks/house_prices" goal="Predict the sales price for each house" eval="RMSE between log-prices"

Things to know

  • Default settings call external LLM APIs, so you need an API key and will pay for model usage unless you run a local model.
  • Fully local runs need an OpenAI-compatible local server and usually show some performance drop compared with hosted models.

In collections

Open-source AI agent frameworks & appsData science courses & tools

Topics

#machine-learning#llm#ai-agents#automated-machine-learning#code-generation#tree-search#data-science#ai#autonomous-agents#autoresearch

FAQ

What is aideml?
An open-source LLM agent that drafts, runs, debugs and iterates on machine learning code via tree search until a user-defined metric is optimized. AIDE ML is the open-source reference implementation of the AIDE algorithm, a tree-search agent that treats each Python script as a node in a solution tree. Given a dataset directory, a goal described in plain language, and an evaluation metric, it generates code, executes it, and uses the metric feedback to decide which candidates to extend or fix.
Is aideml free to use? What license does it have?
aideml is open source under the MIT license. Check the license terms before using it in commercial products.
Is aideml still maintained?
Actively maintained. The last commit to aideml was on Oct 16, 2026.
What can I use aideml for?
Prototyping a baseline ML model on a tabular dataset from a plain-language goal. Replicating or extending the AIDE tree-search experiments in research. Testing alternative search heuristics, evaluators, or LLM backends for ML agents. Running a fully local pipeline against a model served by Ollama.
How do I install aideml?
Install the package from PyPI: pip install -U aideml Set your LLM API key: export OPENAI_API_KEY=<your-key> Run an optimisation on the example task: aide data_dir="example_tasks/house_prices" goal="Predict the sales price for each house" eval="RMSE between log-prices"
Where can I get aideml?
The source code and installation instructions are on GitHub: https://github.com/WecoAI/aideml

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects