MindScribe

AI-powered clinical documentation tool for mental health professionals. MindScribe conducts conversational PHQ-9 and GAD-7 screenings and generates structured clinical notes (DAP, SOAP, BIRP) from therapy session transcripts — powered by Llama 3.3 70B via Groq.

---

Screenshots

Conversational PHQ-9 Screening

Note Generator

---

Features

• Conversational Screening — PHQ-9 and GAD-7 as adaptive dialogue, not a checkbox form; questions adjust tone based on prior answers
• Automatic Scoring — LLM maps free-text patient responses to 0–3 severity scores per item, with heuristic fallback
• Risk Detection — flags suicidal ideation (PHQ-9 Q9) at both keyword and LLM level; always forces a flag on any positive Q9 response
• Note Generation — DAP, SOAP, or BIRP notes from session transcripts in seconds
• ICD-10 & CPT Coding — suggests diagnosis codes and billing codes automatically
• Dashboard — session history, screening trends (Recharts), prior note review
• Demo Mode — fully functional without an API key using pre-seeded data

Quick Start

Backend

cd mindscribe
pip install -r requirements.txt

# Set your Groq API key (get one free at console.groq.com)
cp .env.example .env
# Edit .env and add your GROQ_API_KEY

# Optional: seed demo screening data
python scripts/seed_demo.py

# Start server
python -m uvicorn backend.main:app --reload --port 8000

Frontend

cd frontend
npm install
npm run dev
# Open http://localhost:5173

Demo Mode

Without a GROQ_API_KEY, the app runs in demo mode automatically — adaptive questions use curated static phrasings and notes use realistic pre-written templates. Run python scripts/seed_demo.py first to populate demo screening history.

API Overview

Base URL: http://localhost:8000/api

Method	Endpoint	Description
POST	/screening/start	Start a new PHQ-9 or GAD-7 screening session
POST	/screening/respond	Submit a patient response to the current question
GET	/screening/result/{session_id}	Retrieve completed screening result and score
GET	/screening/history/{patient_id}	Get all past screenings for a patient
POST	/notes/generate	Generate a DAP, SOAP, or BIRP note from a transcript
GET	/clinical/icd10	Search ICD-10 codes (optional ?query= filter)
GET	/demo/transcripts	List bundled demo session transcripts

Evaluation

# Screening accuracy (no backend needed)
python -m evaluation.screening_eval
# Expected: >= 80% pass rate

# Note quality (requires backend running on port 8000)
python -m evaluation.note_eval
# Expected: 100% pass rate

The evaluation suite tests against evaluation/test_transcripts.json — 6 screening cases (PHQ-9 anhedonia severe, PHQ-9 mood moderate, PHQ-9 sleep mild, PHQ-9 suicidal ideation with risk flag, GAD-7 nervousness severe, GAD-7 irritability none) and 4 note generation cases covering all three note formats.

Clinical Disclaimer

MindScribe is a research and demonstration tool only.

• No real patient data should be entered — all data is in-memory and lost on restart
• This system is not HIPAA-compliant infrastructure
• AI-generated notes and scores must be reviewed by a licensed clinician before any clinical use
• PHQ-9 and GAD-7 are validated instruments; this implementation is for workflow demonstration only