mistralai/demo-medical-doc-processor-workflow

mistralai-solutions/demo-medical-doc-processor-workflow

Description: Sample code using Workflows to process various medical documents

Language: MDX

License: Apache-2.0

Stars: 4

Forks: 3

Open issues: 0

Created: 2026-04-22T15:33:22Z

Pushed: 2026-04-27T16:17:32Z

Default branch: main

Fork: no

Archived: no

README:

OCR-driven document processing for healthcare

A Mistral Workflows project.

This project shows how to use Mistral Workflows to process documents in healthcare.

Setup

Run the following command to install the required project dependencies:

make installdeps

Then copy .env.example to .env and set your MISTRAL_API_KEY. You can optionally override the chat completion models via MISTRAL_CLASSIFIER_MODEL and MISTRAL_EXTRACTOR_MODEL.

Commands

Register workflows in AI Studio

Use the following command to auto-discoves all workflow classes in src/workflows/, register them with AI Studio, and starts polling for executions. The task deployent name is set to your hostname:

make start-worker

Execute the workflow

In a separate terminal, start the Streamlit app to view the document extraction UI:

make streamlit

Upload one of the provided PDFs and click "Start Workflow".

You can monitor your workflow's progress and view the extracted data in AI Studio.

Load test

The following command runs a load test that triggers the workflow 10 times using the provided input documents:

make load-test

You can also provide a custom amount of times to trigger the workflow:

make load-test n=5

Development

# Format
uv run ruff format .

# Lint
uv run ruff check --fix .

Clean up

When you're done, stop the Streamlit app and the worker using Ctrl+C.