Session 3: From Analysis to Dashboard
=====================================

Session 3 moves from reusable agent workflows into dashboard delivery. The
central shift is that analysis outputs become a contract between Python work and
the dashboard layer.

Participants examine how the Seattle Public Library teaching dataset was built,
then use percent-cell Python files to generate machine-readable data files,
preserve provenance, and prepare outputs that a browser dashboard can read
directly. Multi-agent workflows are useful here when the parent agent keeps
ownership of the plan while smaller subagents handle bounded investigation or
implementation tasks.

.. csv-table::
   :file: ../csv_tables/session_3_analysis_to_dashboard_pages.csv
   :header-rows: 1

.. toctree::
   :maxdepth: 1
   :hidden:

   self
   building-datasets-with-codex
   multi-agent-workflows
   static-dashboard-prompts
   open-dataset-starting-points