Document Parser

A sample repo using Medplum and AI SDKs for document parsing to the FHIR medical standard.

Setup

This project uses npm for package management. The following packages are installed:

• @medplum/core
• @medplum/fhirtypes
• TypeScript
• OpenAI SDK
• Anthropic SDK
• Vercel AI SDK
• Langchain

Getting Started

To install dependencies:

npm install

Environment Configuration

Create a .env.local by copying the .env.example file and assigning the appropriate environment variables.

Development

This project is set up with TypeScript.

Available Scripts

• npm run create-sample-patient - Runs a script that creates a sample patient.
• npm run document-agent - Runs the document agent script that fetches and processes FHIR documents

Project Structure

• documents/ - the PDF documents
• lib/ - Medplum library
• scripts/ - Utility scripts
  • document-agent.ts - Script for fetching and processing FHIR documents

Document Agent

The document agent script right now only connects to the Medplum API. You are encouraged to use any LLM APIs/SDKs (we have included four of the most common). To run it:

First generate a patient, and take note of its ID:

npm run create-sample-patient

Then request the patient ID when running the document-agent script:

npm run document-agent