health-data. Automated de-identification can miss identifiers; it is not a compliance guarantee on its own.
Who it is forA backend developer who must strip names, addresses and IDs from notes before they reach other systems.
The problemDe-identification inside each app duplicates work and is hard to audit.
The outcomeA local service on port 8080 with /pii/extract and /pii/deidentify endpoints.
What you need first
- Python 3.10+ and pip
- Optional: Apple Silicon for MLX, Android with ONNX Runtime Mobile, or a browser with WebGPU
- Model downloads from the Hugging Face Hub, or local model folders for offline use
Steps
- Run
pip install --upgrade "openmed[hf,service]" - Start it with
python -m openmed.service.loggingplus--host 0.0.0.0 --port 8080 - Check GET /health returns 200 OK
- POST text to /pii/extract to see detected PII, and to /pii/deidentify to get redacted text
- Review a sample of outputs by hand before trusting it in a pipeline
Prompt for your agent
Start the maziyarpanahi/openmed REST service on port 8080 and write a client that sends <synthetic notes> to /pii/deidentify and saves the redacted text. Show me 3 before and after examples.
Caveat. This card is written from the maziyarpanahi/openmed README, not from a test run by Deduped. Before you start, make sure you have Python 3.10+ and pip. Check the repo for newer instructions before you rely on these steps.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for maziyarpanahi/openmed and "Run a local REST API that de-identifies clinical text"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer