🔗 Build a document processing workflow in 30 minutes

Workflows is an orchestration platform for building, executing, and monitoring complex AI-driven workflows. It provides durable, fault-tolerant workflow execution backed by battle-tested distributed systems infrastructure, combined with a developer-friendly SDK.

In this tutorial, you’ll build an end-to-end medical document processing pipeline using three Mistral capabilities: OCR to read PDFs, Agents to classify documents and extract structured data, and Workflows to orchestrate the entire process reliably.

The pipeline takes any scanned medical PDF like a prescription, a hospital bill, or an imaging report and runs it through three steps: optical character recognition (OCR) to extract raw text, an AI agent to classify the document type with a confidence score, and a second agent to extract patient information and document-specific fields as structured JSON. Because it’s built on Mistral Workflows, each step is durable and fault-tolerant: if a worker restarts mid-execution, the workflow resumes where it left off instead of starting over.

You’ll also include a human-in-the-loop step: when the classifier’s confidence falls below a configurable threshold, the pipeline pauses and waits for a user to review and confirm the category before extraction continues. Mistral Workflows enables these long-running processes that can pause and resume based on external input, which is difficult to build reliably with a simple async queue or a chain of API calls.