A multi-agent system for generating realistic clinical case studies for training purposes. The system uses a series of specialized agents to process client information and generate comprehensive case studies with cognitive-behavioral therapy (CBT) elements.

The system consists of four main agents:

1. Preprocessing Agent: Structures raw client input data
2. Case Generator Agent: Enriches the case with cognitive patterns and consequences
3. Narrative Agent: Creates a cohesive first-person narrative
4. Final Review Agent: Combines all outputs into a final case study

• Web-based interface for data input
• Step-by-step case generation process
• JSON-based data structure
• Real-time processing and display
• CBT-focused case formulation

IMPORTANT: Server for 'intermediate' case generator work on 3001 port, base case generator work on 3000 port.

1. Clone the repository and go to the backend folder:

   cd backend

2. Install dependencies:

   pnpm install

3. Create a .env file in the root directory with your OpenAI API key:

   ANTHROPIC_API_KEY=your_api_key_here
   

1. Start the server:

   node server.js (3000 port)
   node case_generator_server.js (3001 port)

2. Open frontend/case_generator.html in your browser

i recomend use httpserver for frontend:

npx http-server

3. Fill in the client information form and submit to generate a case study

• POST /api/preprocess: Processes raw client data
• POST /api/generate-case: Generates enriched case data
• POST /api/generate-narrative: Creates narrative description
• POST /api/finalize-case: Produces final case study

{
  "client_profile": {
    "age": 35,
    "gender": "Female",
    "occupation": "Software Engineer"
  },
  "presenting_problem": {
    "symptoms": [
      "persistent sadness",
      "loss of interest",
      "fatigue"
    ],
    "trigger_event": "Recent breakup with a long-term partner"
  }
}

{
  "client_profile": { ... },
  "presenting_problem": { ... },
  "cognitive_patterns": {
    "automatic_thoughts": [ ... ],
    "core_beliefs": [ ... ]
  },
  "consequences": {
    "emotional": [ ... ],
    "behavioral": [ ... ]
  },
  "client_narrative": "..."
}

• The system requires an OpenAI API key
• CORS is enabled for local development
• Input validation is performed at multiple stages
• Sensitive clinical data should be handled according to relevant privacy regulations

The project uses:

• Express.js for the backend
• Anthropic Claude Haiku for text generation
• Bootstrap for the frontend UI
• Vanilla JavaScript for frontend logic

MIT License