Skip to content

Notebook UI

hiperhealth.notebook provides an optional Jupyter widget interface for the session-backed clinical pipeline workflow.

Install the optional dependencies:

pip install "hiperhealth[notebook]"

Launch the UI from a notebook:

from hiperhealth.notebook import ui

ui.show()
ui.show(data_dir="/my/data/path")

The UI follows the workflow from docs/example.qmd:

  1. create or load a parquet-backed session
  2. enter de-identified screening data
  3. run screening and intake
  4. check diagnosis requirements, answer inquiries, and run diagnosis
  5. run exam suggestions
  6. reload the same session after lab results arrive
  7. rerun enriched diagnosis
  8. run treatment and prescription stages
  9. inspect results and the session event log

Skill app views

Skills may declare optional app views in their skill.yaml metadata. The notebook UI discovers active skill views and renders them on the Skill apps page. The declaration uses JSON Schema-style data_schema plus JSON Forms-style ui_schema, so the same metadata can be reused by future notebook or JupyterLab renderers.

Values saved through skill app views are persisted through session events and replayed into PipelineContext.extras['skill_ui'] for the skill to consume.

Session files may contain sensitive clinical artifacts. Protect the configured data_dir like clinical records, do not store API keys in it, and prefer synthetic or de-identified data in examples.