▶️ AGILAB (Home dashboard)
==============================

AGILAB.py is the main Streamlit entry point. It provides a single navigation
surface for the built-in pages (EDIT / EXECUTE / EXPERIMENT / EXPLORE) and for
optional app-specific dashboards ("apps-pages") that can be enabled per project.

Navigation model
----------------

- **Built-in pages** are always available and drive reproducible workflows:

  - **EDIT**: inspect and modify the project sources and settings.
  - **EXECUTE**: install workers, generate distributions, and run pipelines.
  - **EXPERIMENT**: iterate on analysis/code in `lab_steps.toml` and run snippets
    against exported datasets.
  - **EXPLORE**: browse and launch installed Streamlit bundles (apps-pages).

- **Apps-pages** are standalone Streamlit bundles discovered under
  ``${AGILAB_PAGES_ABS}``. They run in isolated virtual environments and are
  enabled by listing their module names under ``[pages].view_module`` in
  ``app_settings.toml``.

Learning lifecycle (when applicable)
------------------------------------

When a project includes learning components:

- **Training** updates model parameters using new experience.
- **Inference** runs a fixed checkpoint to produce allocations/decisions.
- **Continuous learning** and **federated learning** require explicit pipeline
  steps (checkpointing, dataset joins, aggregation); AGILAB provides the run
  orchestration and artifact conventions, but does not enable these modes
  implicitly.

See also
--------

- :doc:`architecture` for the end-to-end pipeline view.
- :doc:`execute-help` and :doc:`experiment-help` for the default workflow.
- :doc:`learning-workflows` for training vs inference, continuous learning, and
  federated learning patterns.