Telepathy is a Next.js healthcare web app with patient and clinician portals, AI-assisted documentation, multilingual support, and 3D body-map workflows.

• Role-based auth (patient/clinician) with signup and login.
• Dashboard, appointments, current visit, visit history, profile, and report pages.
• Find-doctor flow with assignment and clinician browsing.
• 3D body map to mark pain regions and add notes.
• Upload media during current visit (images/video) for clinician review.
• Visit summary and historical records view.

• Clinician dashboard, appointments, patient list, current visit, records, and body-map pages.
• Live-style current visit workspace with transcript feed, EMR form autofill, priority markers, and diagnosis visualizations.
• Doctor and AI views over patient-submitted body regions.
• Patient detail views with records and contextual data.

• EMR extraction from visit conversation.
• Financial and clinical insights from patient profile data.
• Possible-differential generation based on patient/visit context.
• Translation support for multilingual workflows.
• ElevenLabs STT/TTS integration for voice input/output.
• Media/frame analysis endpoints for visit artifacts.

This project uses Featherless as an OpenAI-compatible LLM backend.

• Wrapper: lib/featherless.ts
• AI orchestration: lib/aiService.ts
• Base URL: https://api.featherless.ai/v1
• Required key: FEATHERLESS_API_KEY
• Primary models used:
• Qwen/Qwen3-32B for general analysis/translation workflows.
• nicoboss/Qwen-3-32B-Medical-Reasoning for structured medical note generation.
• Request behavior:
• retries on transient errors (e.g., 429/503),
• disables "thinking" output where supported,
• strips <think>...</think> blocks before parsing.

Gemini can be used as a supplemental provider for EMR and differential generation when GEMINI_API_KEY is configured (lib/gemini.ts).

Transcription uses ElevenLabs Speech-to-Text (Scribe v2).

Flow:

1. Patient records a voice note in current visit and chooses language (optional).
2. On submit, app uploads media to POST /api/visits/[id]/media, then calls POST /api/visits/[id]/submit with { speechLanguage }.
3. Submit marks visit as submitted, then triggers voice processing via POST /api/visits/[id]/process-voice.
4. Voice processing reads patientMedia.voiceRecordingUrl, sends audio to ElevenLabs STT, optionally translates to English, and saves transcript into visit.conversation.

Requirements:

• ELEVENLABS_API_KEY in .env.local.
• Voice must be uploaded before submit.
• Supported input: audio data URL (e.g. audio/webm), up to 5MB per recording.

If transcript is missing:

• Verify ELEVENLABS_API_KEY is valid.
• Check server logs for ElevenLabs STT API error or Process-voice STT failed.

• Framework: Next.js 14 (App Router), React 18, TypeScript
• Styling/UI: Tailwind CSS, Framer Motion, Lucide React
• Data layer: MongoDB + Mongoose
• 3D body-map: Three.js + @react-three/fiber + @react-three/drei
• AI/LLM:
• Featherless (OpenAI SDK compatibility)
• Google Gemini (@google/genai)
• Voice:
• ElevenLabs client + SDK (@elevenlabs/client, @elevenlabs/elevenlabs-js)
• Charts/visuals: Recharts
• Auth/security primitives: bcryptjs for password hashing

• app/ - App Router pages and API routes
• components/ - shared UI and feature components
• components/body-views/ - 3D body model and body-map views
• lib/ - DB, AI, model schemas, utility modules
• public/ - static assets (including body-model files and logo assets)

Main route groups under app/api:

• Auth: /api/auth
• Seed data: /api/seed
• Patients: /api/patients, /api/patients/[id], /api/patients/[id]/visits, /api/patients/[id]/records
• Clinicians: /api/clinicians, /api/clinicians/[id], /api/clinicians/[id]/patients
• Visits: /api/visits, /api/visits/[id], /api/visits/invite
• Visit AI/media/body-map:
• /api/visits/[id]/submit
• /api/visits/[id]/process-voice
• /api/visits/[id]/possible-diseases
• /api/visits/[id]/body-map
• /api/visits/[id]/media
• /api/visits/[id]/analyze-media
• AI/utilities:
• /api/pipeline, /api/translate, /api/transcribe, /api/tts
• ElevenLabs-specific:
• /api/elevenlabs, /api/elevenlabs/stt

Create .env.local in repo root:

MONGODB_URI=your_mongodb_connection_string
FEATHERLESS_API_KEY=your_featherless_api_key
GEMINI_API_KEY=your_gemini_api_key_optional
ELEVENLABS_API_KEY=your_elevenlabs_api_key_optional

Notes:

• MONGODB_URI is required for app startup.
• Featherless/Gemini/ElevenLabs keys enable AI and voice features.

Install dependencies:

npm install

Run dev server (default script binds to 127.0.0.1:3000):

npm run dev

If port 3000 is busy, override:

npm run dev -- --port 3020 --hostname 127.0.0.1

Populate demo clinicians, patients, visits, and records:

curl -X POST http://127.0.0.1:3000/api/seed

Use test accounts listed in:

• test-accounts.md

• npm run dev - start local dev server
• npm run build - production build
• npm run start - run production server
• npm run lint - run Next lint

• Next.js version in this repo is 14.2.25 (npm may warn about security patches; upgrade when feasible).
• Some AI/voice features gracefully degrade when optional API keys are missing.