Clinical Note Tool (Next.js + MongoDB + GPT)

This app ingests clinical documents, structures key clinical fields, generates an HPI narrative, and stores a versioned HPI history per case. It also supports importing MCG-style criteria documents (mapped by disease/pathway) separately from cases.

Basic Functions (User Instructions)

Cases (clinical documents + HPI workflow)

Open /cases.

1. Create a new case: /cases/new.
2. Upload clinical documents (PDF or DOCX) for the case.
3. The system auto-classifies document type (ER_NOTE / HP_NOTE / OTHER), extracts text, and produces structuredRawData.
4. In the case page, click Generate HPI to create an HPI and append it to the generatedHPI history.
5. For each generated HPI entry:
   • Review runs a payer/UM score + missing points + inconsistencies + improvements.
   • Regenerate review regenerates a new HPI using the review output.
   • Regenerate HPIbased on the review, user can select the point at to the prompt, and also manually input, to regenerate a new version of HPI.
   • Edit allows direct modification with a simple choice:
     • Update this HPI: modifies the current entry, sets type: human_revise, and clears score/review fields.
     • Save as new HPI: creates a new entry with type: human_revise and clears score/review fields for the new one.
   • Delete removes the entry from history.
6. Auto generate loop runs a hybrid loop (generate 2 candidates, review both, then refine up to 10 times with early stopping). It saves only the final best-scoring HPI once.

MCG criteria (separate from cases)

Open /mcg.

1. Upload an MCG guideline PDF or DOCX.
2. The system extracts a structured JSON criteria map keyed by disease/pathway identifiers.
3. View the extracted criteria in /mcg/[id].

Setup

Prerequisites

• Node.js (Node 20+ recommended)
• MongoDB connection string
• OpenAI API key

Environment variables

Create .env.local with:

MONGODB_URI=your_mongodb_connection_string
OPENAI_API_KEY=your_openai_key

Local dev

npm install
npm run dev

Then open http://localhost:3000.

Architecture Overview

Frontend

• Next.js App Router under src/app/
• Case UI:
  • src/app/cases/[id]/page.tsx: loads case detail and renders main sections
  • src/app/cases/[id]/merged-hpi-summary.tsx: HPI generation/review/regenerate/edit UI

Backend (API routes)

• src/app/api/cases/[id]/ingest/route.ts: upload PDFs/DOCX and structure into structuredRawData
• src/app/api/cases/[id]/generate-hpi/route.ts: generate HPI and append to generatedHPI
• src/app/api/cases/[id]/generated-hpi/route.ts:
  • DELETE deletes an HPI entry
  • PATCH edits an HPI entry (update current or save new)
• src/app/api/cases/[id]/generated-hpi/review/route.ts: payer/UM review stored as score + improvement
• src/app/api/cases/[id]/regenerate-hpi/route.ts: regenerates using review instructions
• src/app/api/cases/[id]/auto-generate-hpi-loop/route.ts: server-side loop; saves only the final HPI

Core libraries

• HPI generation: src/lib/generate-hpi-from-summary-gpt.ts
• Review: src/lib/generate-hpi-review-gpt.ts
• Regeneration: src/lib/generate-hpi-regenerate-gpt.ts
• Auto-loop orchestration (in-memory): src/lib/auto-hpi-loop.ts
• MongoDB persistence: src/lib/cases-db.ts

MCG extraction

• src/lib/structure-mcg-gpt.ts and src/lib/mcg-db.ts
• UI: src/app/mcg/page.tsx and src/app/mcg/[id]/page.tsx

Deploy to AWS Amplify

1) Connect your repo

1. AWS Amplify Console -> Create app
2. Connect Git provider (GitHub, GitLab, etc.)
3. Select branch (e.g. main)

2) Build settings (typical for Next.js)

Set:

• Build command: npm run build
• Start command: npm run start

Amplify usually detects Next.js automatically; if not, choose the Next.js framework preset.

3) Configure environment variables in Amplify

Add:

• MONGODB_URI
• OPENAI_API_KEY

Then redeploy.