HL7 FHIR Implementation Guide Analyzer

This project contains the plain language summaries for published HL7 implementation guides, and also AI based scripts to fetch, analyze, and summarize HL7 FHIR Implementation Guides (IGs).

Plain Language Summaries

• The plain language summaries are maintained in this repository: https://github.com/HL7/plain-language
• The summary for each IG is stored in summaries/{id}.md, where {id} is the package id for the implementation guide
• There may be candidate plain language summaries produced by in place already
• The candidate summaries are ignored by hl7.org and fhir.org unless they start with the line # General
• Whatever is under # General is the version independent plain language summary for the IG (up to the next line starting with "# ")
• If you want version specific plain language summaries, add additional summaries by adding a line # {ver} where ver is the version.
• Make a PR, and one of the exec team can approve it.

As an example, https://fhir.org/guides/acme/HotBeverage/1.4.2022/ has a plain language summary:

This is authored by the file summaries/fhir.acme.hotbeverage.md:

# General

This (fake) standard was published on 1st April 2022 as an insider's April Fools Joke. No further versions are planned (unless/until someone else *likes* the joke)

# 1.4.2022

This (fake) version of the standard was published on 1st April 2022 as an insider's April Fools Joke. No further versions are planned (unless/until someone else *likes* the joke)

If it matters, the markdown syntax is commonmark

AI Scripts

Prerequisites

• bun (latest version)
• Git
• Google Cloud SDK (for Vertex AI)
• Anthropic API key (for Claude)

Setup

1. Clone this repository:

   git clone <repository-url>
   cd <repository-name>
   

2. Install dependencies:

   bun install
   

3. Set up Google Cloud credentials:

   gcloud auth application-default login
   

4. Set your Anthropic API key as an environment variable:

   export ANTHROPIC_API_KEY=your_api_key_here
   

Usage

1. Fetch HL7 Repositories

Run the hl7repo.ts script to fetch HL7 repositories:

bun run hl7repo.ts

This script will:

• Fetch a list of HL7 repositories from https://build.fhir.org/ig/HL7/
• Create a repos directory
• Shallow clone each repository into the repos directory

2. Process Repositories

After fetching the repositories, you can process them using the process_repo.ts script:

bun run process_repo.ts repos/<repository-name>

Replace <repository-name> with the name of the repository you want to process.

This script will:

• Concatenate relevant files from the repository's input directory
• Generate an initial analysis using Google's Vertex AI
• Refine the analysis using Anthropic's Claude AI
• Save the results in the following directories:
  • prompts: Contains the prompts used for AI analysis
  • analysis: Contains the initial AI analysis
  • summaries: Contains the final refined summaries

Output

• repos/: Contains the cloned HL7 repositories
• prompts/: Contains the prompts used for AI analysis
• analysis/: Contains the initial AI analysis for each IG
• summaries/: Contains the final refined summaries for each IG

Each summary in the summaries/ directory is a markdown file named after the IG, containing a concise, plain-language description of the IG's purpose, features, and relevance.

Notes

• The scripts use AI models, so results will vary between runs.
• Ensure you have sufficient permissions and API quota for Google Cloud and Anthropic services.
• Processing large repositories may take some time and consume API resources.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

The code in this project is licensed under the MIT License. The summaries themselves are Creative Commons Public Domain.