Usage: Web UI (Streamlit)

AIDE ML includes a simple and intuitive web interface built with Streamlit for users who prefer a graphical interface over the command line.

1. Installation

To use the Web UI, you need to install streamlit along with aideml. If you installed aideml via pip, Streamlit is included by default.

pip install -U aideml

2. Running the App

Navigate to the webui directory within the aide package and run the Streamlit app.

# Find the installation path of the aideml package
# Then navigate to its webui directory
cd $(pip show aideml | grep Location | cut -d ' ' -f 2)/aide/webui

# Run the streamlit app
streamlit run app.py

Your web browser should open with the AIDE ML interface loaded.

3. Using the UI

The interface is designed to be straightforward:

  1. Settings Sidebar: On the left, there is a sidebar for configuration.

    • API Keys: Paste your required API keys (e.g., OpenAI, Anthropic) into the respective fields and click "Save API Keys". These are stored in the session state for the duration of your browser session.

  2. Input Section: The main panel on the left is for your task inputs.

    • Data Files: You can either drag and drop your dataset files (CSV, TXT, etc.) into the file uploader or click the "Load Example Experiment" button to use the built-in House Prices dataset.
    • Goal: Describe your machine learning objective in the "Goal" text area.
    • Evaluation Criteria: Specify the metric for success in the "Evaluation Criteria" text area.
    • Number of Steps: Use the slider to select how many iterative steps the agent should perform.

  3. Run AIDE: Once your data and instructions are set, click the "Run AIDE" button.

  4. Results Section: The right-hand panel will come to life, showing live progress and results.

    • A progress bar will show the current step.
    • Tabs will appear allowing you to inspect the results:
      • Tree Visualization: An interactive plot of the solution tree.
      • Best Solution: The code for the best-performing solution found so far.
      • Config: The full YAML configuration for the run.
      • Journal: A raw JSON log of every action the agent takes.
      • Validation Plot: A live plot of the validation score over time.