lrscribe76-api

Go backend for LRScribe76 — a medical scribe app that records audio, transcribes it, and generates structured clinical documents using LLMs.

Handles audio recording (proxied to a dedicated audio service), transcription via Gemini, document generation via Requesty/OpenAI, and CRUD for notes and transcriptions backed by Postgres.

Quick Start

# Prerequisites: Go 1.21+, PostgreSQL

git clone https://github.com/leonj1/lrscribe76-api.git
cd lrscribe76-api

# Set required env vars (see below)
export DATABASE_URL="postgres://user:pass@localhost:5432/lrscribe76"
export CLERK_SECRET_KEY="sk_test_..."
export REQUESTY_API_KEY="rqsty-sk-..."
export AUDIO_API_URL="https://lrscribe-audio-api-production.up.railway.app"
export AUDIO_API_KEY="your-audio-api-key"

go run notes.go
# Server starts on :8080

Railway Deployment

Field	Value
Project	lrscribe-audio-api
Service	lrscribe76-api
Environment	production
Public Domain	https://lrscribe76-api-production.up.railway.app
Private Domain	lrscribe76-api.railway.internal

Sibling Services (same Railway project)

Service	Domain	Description
lrscribe76	lrscribe76-production.up.railway.app	Next.js frontend + Express middleware
lrscribe-audio-api	lrscribe-audio-api-production.up.railway.app	Audio chunk storage & S3 management

Environment Variables

Variable	Required	Default	Description
DATABASE_URL	Yes	—	PostgreSQL connection string
CLERK_SECRET_KEY	Yes	—	Clerk authentication secret key
REQUESTY_API_KEY	Yes	—	API key for Requesty (LLM router)
REQUESTY_MODEL	No	openai-responses/gpt-5.4	LLM model for document generation
AUDIO_API_URL	Yes	—	Base URL of the audio recording service
AUDIO_API_KEY	Yes	—	Auth key for the audio service
VITE_CONVEX_URL	Yes	—	Convex deployment URL (for session/user data)
PORT	No	8080	Server listen port

API Endpoints

Health

• GET /health — Service health check

Auth

• GET /api/auth/user — Current user info (Clerk JWT)

Document Generation

• POST /api/generate-document — Generate structured clinical document from transcription
• POST /api/regenerate-section — Regenerate a single document section

Transcription

• POST /api/transcribe — Transcribe base64-encoded audio via Gemini
• POST /api/transcribe-from-url — Transcribe audio from URL or recording ID (Clerk JWT)

Audio Recording (Convex auth)

• POST /api/audio/start — Start a recording session
• POST /api/audio/chunk/:recordingId — Upload audio chunk
• GET /api/audio/status/:recordingId — Recording status
• POST /api/audio/complete/:recordingId — Finalize recording
• POST /api/audio/trigger-interim/:recordingId — Trigger interim transcription, start new segment

Notes & Transcriptions (requires Postgres)

• GET /api/transcriptions — List transcriptions
• POST /api/transcriptions — Create transcription
• GET /api/transcriptions/:id — Get transcription (Clerk JWT)

Full endpoint documentation with request/response examples: ENDPOINTS.md

Tech Stack

• Go 1.21 with vestigo router
• Clerk for JWT authentication
• PostgreSQL (via lib/pq) for notes & transcriptions
• Convex for real-time session/user data
• Requesty as LLM router (OpenAI, Gemini models)
• S3 for audio storage (via audio service)

License

Private repository.