HiPerHealth¶

A Python library for clinical AI workflows.
Composable, stage-independent pipelines for screening, diagnosis, treatment, and more.

[Get Started](installation.md){ .md-button .md-button--primary } [View on GitHub](https://github.com/hiperhealth/hiperhealth){ .md-button }

Skill-Based Pipeline¶

Stages run independently, at different times, by different actors. Compose clinical workflows from modular skills.

Session Files¶

Parquet-backed event logs for persistent, resumable clinical workflows across multiple patient visits.

Requirement Checking¶

Skills declare what information they need before execution, with three priority levels: required, supplementary, and deferred.

Built-in Skills¶

DiagnosticsSkill for differential diagnosis, ExtractionSkill for PDF/image reports, and PrivacySkill for PII de-identification.

Extensible¶

Create custom skills as Python classes, install third-party skills via entry points or Git URLs.

Data Science Friendly¶

Sessions are standard parquet files, queryable with pandas, polars, or DuckDB. Use from Jupyter notebooks.

Quick Start¶

Install from PyPI:

pip install hiperhealth

Run a diagnosis:

from hiperhealth.pipeline import PipelineContext, Stage, create_default_runner

runner = create_default_runner()

ctx = PipelineContext(
    patient={"symptoms": "chest pain, shortness of breath", "age": 45},
    language="en",
    session_id="visit-1",
)

ctx = runner.run(Stage.DIAGNOSIS, ctx)
print(ctx.results["diagnosis"].summary)

Documentation Guide¶

Section	Description
Installation	Install hiperhealth and system dependencies
LLM Configuration	Configure LLM backends (OpenAI, Ollama, Groq, and more)
Usage	End-to-end examples: pipeline, sessions, extraction
Creating Skills	Build and register custom pipeline skills
API Reference	Auto-generated Python API documentation
Changelog	Release notes and version history
Contributing	Development setup and contributor guide

Python