Skip to main content
ClaimJumper runs as a standard Next.js application. You need Node.js >=22 <23, npm, and an OpenAI API key with access to the models used by the pipeline. The repository includes a .nvmrc file so version managers like nvm and fnm can pin the correct Node version automatically.

​
Requirements

  • Node.js >=22 <23 — use the included .nvmrc with nvm (nvm use) or fnm (fnm use) to select the right version.
  • npm — bundled with Node.js; no separate installation needed.
  • An OpenAI API key with access to gpt-5.6-sol (used for ingest and triage) and gpt-5.6-terra (used for drafting).

​
Setup and first run

1

Clone the repository

2

Install dependencies

This installs Next.js, React, the OpenAI SDK, Zod, Vitest, and all other project dependencies listed in package.json.
3

Configure your API key

Copy the example environment file and add your key:
Then open .env.local and replace the placeholder with your real key:
.env.local
The application reads OPENAI_API_KEY server-side only. Never commit .env.local.
4

Start the development server

Next.js will print the local URL to the terminal (typically http://localhost:3000). Open that URL in your browser.
5

Upload the sample EOB

The repository includes samples/eob-001.png, a demo remittance screenshot. Upload it through the UI to see all five routing lanes in action: appeal, corrected claim, patient bill, write-off, and human review. Parsed output is cached to samples/parsed/ after the first run, so subsequent uploads of the same file skip the ingest API call.

​
Verification commands

Run the full verification suite before opening a pull request or declaring a change complete:
The test suite uses injected dependencies throughout the pipeline. Live model calls are not made during npm test — tests that would require real API responses are tagged with Vitest’s skip to keep the suite fast and reproducible without a valid key. You can run the full suite offline.

​
Development scripts

All scripts are defined in package.json:

​
Sample data

samples/eob-001.png is a demo remittance screenshot included in the repository. It is used by the walkthrough and referenced in the regression test fixtures. The image exercises all five routing lanes, making it suitable for a repeatable local demo. Parsed ingest output for this file is cached at samples/parsed/ after the first API call. The samples/parsed/ directory is writable at runtime; you can delete its contents to force a fresh ingest API call on the next upload. eob-001.png is demo data only. Do not treat it as production patient data, and do not substitute real patient records for local testing without appropriate access controls in place.