Rare-disease support often begins before a diagnosis: fragmented symptom descriptions need to be organized into terminology that specialists can review. This Astron for Good application concept shows how an Astron Agent workflow can structure that preparation while preserving an audit trail.
The demo uses synthetic or de-identified clinical text. It does not provide diagnosis, treatment advice, or patient-triage decisions.
What the workflow does
The importable Astron DSL executes this chain:
- Accept synthetic or de-identified clinical text.
- Call a configurable local bridge through
OPENRARE_BRIDGE_BASE_URL. - Run OpenRare-compatible phenotype extraction.
- Constrain term mapping with the CHPO dictionary.
- Format HPO identifiers, source commit, provider model, elapsed time and trace.
- Return a structured result for professional review.
The local run used the fixed synthetic example of polydactyly, atrial septal defect and growth delay. It returned three auditable clues: HP:0001161, HP:0001631 and HP:0001510. The point is not to infer a diagnosis; it is to make the clues easier to verify and hand over.
Why this is an Astron for Good scenario
Patient organizations and public-interest volunteers often help people collect documents, normalize terminology and prepare questions for professional appointments. A workflow can reduce repetitive formatting, keep source and run metadata together, and make the hand-off more consistent. Human review remains the center of the process.
Reuse the workflow
- Workflow hub: https://www.awesome-astron-workflow.dev/#workflows
- Chinese demo: https://youtube.com/shorts/7QxpK6pvSNU
- English demo: https://youtube.com/shorts/n-_oQ_yB0nk
- OpenRare: https://github.com/OpenRare2026/OpenRare
- Astron Agent: https://github.com/iflytek/astron-agent
The workflow files are provided in Chinese and English. Configure the local bridge endpoint in the Astron workflow runtime before importing and running them.
Top comments (0)