MedCaseAgent

MedCaseAgent

A project page for research resources on rare disease clinical case reports: a 584-record benchmark, a large corpus of case-report metadata and the agent code. For medical AI researchers.

Made by PediaMedAI on Hugging Face. These notes are Mindket's, written from its Space page and README; the app's own files and license are on Hugging Face.

About

What it is: MedCaseAgent is a project page that gathers three research resources on clinical case reports, with a focus on rare diseases. MedCase-Bench holds 584 structured clinical input records for judging case-report generation. MedCase-150K is a large corpus of case-report metadata. The third is the MedCaseAgent code. The page itself is a static site.

Who it suits: It suits researchers who work on generating clinical case reports from medical evidence, and students who want to see how such a project is organised.

Title: Who might use this project page
Industry | Who uses it | How they use it | Example
Medical AI research | Researchers | Find the benchmark, corpus and code in one place | Follow the page's links to the benchmark records
Natural language processing | Engineers | Study how a report-generation agent is built | Read the code's planner, writer and editor stages
Medical education | Students | See how clinical evidence becomes a case report | Look at the teaser figure on the page

If you want a tool for real patients, you can skip it, since this is research material and the page accepts no patient uploads.

Where to use it: You can read it online or keep a copy of its files.

• Web browser: the Space's page on Hugging Face, with nothing to install
• Windows, macOS or Linux: the downloaded files, opened in a browser
• Node.js 22.12 or newer: only if you want to rebuild the page from its source

Title: What the project page is good for
Use case | Who | What they get back
Find the research resources | Researchers | One page that points to the benchmark, the corpus and the code
Understand the project | Students | An introduction and a teaser figure that work in the browser
Rebuild or update the page | Maintainers | A static site rebuilt from the repository's source

How to set it up: The hosted page needs nothing. To get the files you need Python and pip.

• Python, with pip
• Any web browser
• Node.js 22.12 or newer, only for rebuilding

1. Install the huggingface_hub package.
2. Download the Space's files into a folder.
3. Move into that folder.
4. Open index.html in a web browser.

pip install -U huggingface_hub
hf download PediaMedAI/MedCaseAgent --repo-type space --local-dir MedCaseAgent
cd MedCaseAgent

To rebuild the static front end from source, run these inside the project.

cd frontend
npm ci --include=dev
npm run build:static

How to use it: Open the Space online, or open index.html from your downloaded folder. The introduction, the teaser figure and the links to the resources all work in the browser. Maintainers who change the public content edit one Python file and run an export.

python -m venv .venv
.venv/bin/pip install -r backend/requirements.txt
.venv/bin/python scripts/export_static.py
cd frontend
npm run build:static

Title: The resources the page brings together
Tool | What it does
MedCase-Bench | A set of 584 clinical input records, organised for testing how well a model writes case reports
MedCase-150K | A large corpus of case-report metadata, including rare disease reports
MedCaseAgent code | Preprocessing, planner, writer and editor report-generation inference, and a disease retrieval index

Example: A researcher opens the project page, reads the introduction and looks at the teaser figure. Then they follow the link to MedCase-Bench to find the 584 structured clinical records they can use to evaluate report generation.

For newcomers and experts: If you are new to this, open the Space and read the introduction first. This is a research project page, and it is not a clinical tool.

If you know the field, the page is a React, TypeScript and Vite front end that reads public JSON snapshots exported from an optional FastAPI backend. The built files are compiled locally and uploaded to the Space root next to the source, so no hosted build runs. The notes warn against turning on a build command there, since hosted builds may cost account credits.

Edit the content in backend/app/content.py
Export the validated public JSON data
Build the static front end
Upload frontend/dist to the Space root

Good to know: The page links to resources and does not host them.

• Scope: the page leaves the datasets where they are, stores no model weights and takes no patient uploads
• Static hosting: nothing runs on a server, and no model, token or paid hardware is involved
• Docker: a Docker Space may need a paid Hugging Face plan, and the published static Space does not use one
• FastAPI mode: the full backend stays available as an option for future extensions
• License: cc-by-nc-4, and the code's license restricts commercial use while the source publications keep their own licenses
• Cost to run: none named for the static page, and Node.js 22.12 or newer for rebuilding
• Data it reaches: public JSON snapshots, and exports never include environment variables or credentials
• Not for: commercial use, because the code license forbids it

On Hugging Face, 13 November 2026

• Likes: 1
• Built with: static
• License: cc-by-nc-4.0

Links

• Space: https://huggingface.co/spaces/PediaMedAI/MedCaseAgent
• Files: https://huggingface.co/spaces/PediaMedAI/MedCaseAgent/tree/main

Try it

Open the Space in your browser, or copy it to your own Hugging Face account from its page. Nothing in this kit runs by itself.