▶️ EXPERIMENT

Main Content Area

Each lab is organised as a sequence of steps stored in lab_steps.toml. The numbered buttons at the top let you jump between them. Ask questions or describe transformations in the text area—AGILab forwards the prompt to the Responses API together with the selected DataFrame metadata. The code editor reacts to the toolbar actions:

  • Save keeps the snippet as-is in the current step.

  • Next persists the snippet and advances to a fresh step.

  • Remove deletes the step from lab_steps.toml.

  • Run writes the snippet to lab_snippet.py, executes it and stores any produced dataframe under lab_export_out.csv so the preview and the Execute/Explore pages can consume the result.

The runtime is chosen from the Execution environment box below the editor. If you pick a concrete virtual environment path the snippet runs via run_agi inside that environment (the path is kept with the step under the E field). Leaving the selector on the default AGILab environment falls back to run_lab, reusing the managed runtime that ships with the app. In both cases the exported dataframe and history behave identically.

The assistant automatically reloads the most recent dataframe and shows it below the editor. If nothing has been saved yet, you will see a reminder to run a snippet first.

Inspect or tweak the raw lab_steps.toml via the code editor. Saving the file here immediately refreshes the assistant tab.