Launchpad Scribe

A web application for transcribing audio and video files into polished Word documents. Built with SvelteKit (frontend) and FastAPI (backend), packaged as a single Docker container.

Migration note: This project was previously a Tauri/Rust desktop application. It has been fully converted to a web app using the OHG Launchpad solution template.

Features

Feature	Description
Drag & Drop	Drop audio/video files to transcribe
AI Transcription	Uses AssemblyAI for high-quality speech-to-text
Speaker Diarization	Automatically identifies and labels different speakers
Word Export	Generates professional .docx documents
Vocabulary Lists	Custom and system vocabulary for domain-specific terms
Topic Detection	Optional AI-powered topic extraction
Sentiment Analysis	Optional per-utterance sentiment
Transcription History	Browse and re-export past transcriptions
Meeting Minutes	AI-powered minutes generation via Anthropic Claude

---

Supported Formats

AssemblyAI accepts audio and video files directly (up to 5GB).

Type	Extensions
Video	.mp4, .mov, .avi, .mkv, .webm
Audio	.mp3, .wav, .m4a, .aac, .ogg, .flac

---

Architecture

launchpad-scribe/
├── frontend/                     # SvelteKit web app
│   └── src/
│       ├── routes/               # Pages
│       │   ├── +page.svelte      # Main transcription queue
│       │   └── transcript/[id]/  # Transcript viewer
│       └── lib/
│           ├── components/       # UI components
│           ├── services/         # API & business logic
│           │   ├── transcription.ts  # AssemblyAI integration
│           │   ├── docx-export.ts    # Word document generation
│           │   └── history.ts        # Transcript history
│           └── stores/           # Svelte state stores
│
├── backend/                      # FastAPI Python API
│   └── app/
│       ├── api/                  # Route handlers
│       │   ├── transcription.py  # AssemblyAI proxy + polling
│       │   ├── ai.py             # Anthropic minutes generation
│       │   ├── history.py        # Transcript history CRUD
│       │   ├── presets.py        # Transcription preset CRUD
│       │   ├── vocabularies.py   # Vocabulary CRUD + extraction
│       │   └── jobs.py           # Background job tracking
│       ├── services/
│       │   ├── assemblyai.py     # AssemblyAI API client
│       │   └── anthropic.py      # Anthropic API client
│       └── db/                   # SQLAlchemy models + Alembic
│
├── docker-compose.yml            # App + PostgreSQL
├── Dockerfile                    # Multi-stage build (Node → Python)
└── backend/.env                  # Local environment variables

---

Key Technologies

Layer	Technology
Frontend	SvelteKit 2 + TypeScript
Backend	FastAPI (Python 3.11)
Database	PostgreSQL + SQLAlchemy + Alembic
Transcription	AssemblyAI API
AI Minutes	Anthropic Claude (via API)
Container	Docker (single container, multi-stage build)
Word Export	docx npm package (client-side)

---

Data Flow

graph LR
    A[Drop File] --> B[XHR Upload to AssemblyAI]
    B --> C[POST /api/transcription/submit]
    C --> D[Poll /api/transcription/:id every 3s]
    D --> E[Parse Response]
    E --> F[Generate .docx - auto download]
    E --> G[POST /api/history - save to DB]
    G --> H[Appears in History sidebar]

Loading

---

Local Development

Prerequisites

• Docker Desktop
• AssemblyAI API key
• Anthropic API key

Setup

1. Create backend/.env:

DATABASE_URL=postgresql+psycopg://launchpadscribe:launchpadscribe@db:5432/launchpadscribe
ASSEMBLYAI_API_KEY=your_key_here
ANTHROPIC_API_KEY=your_key_here
DEV_USER_EMAIL=you@example.com

2. Start the stack:

docker-compose up --build

3. Open: http://localhost:8000

The container runs Alembic migrations automatically on startup, then serves the SvelteKit SPA and FastAPI on the same port.

---

Environment Variables

Variable	Required	Description
DATABASE_URL	✅	PostgreSQL connection string (use postgresql+psycopg://)
ASSEMBLYAI_API_KEY	✅	Server-managed — never sent to clients
ANTHROPIC_API_KEY	✅	Used for minutes generation
DEV_USER_EMAIL	Dev only	Bypasses Entra ID auth for local dev
STATIC_DIR	Auto-set	Path to SvelteKit build output

---

Authentication

• Production: Azure Entra ID via x-ms-client-principal-name header (set by Azure Container Apps)
• Development: Set DEV_USER_EMAIL in .env to bypass auth

---

Database

PostgreSQL with Alembic migrations. Key tables:

Table	Purpose
users	Auto-provisioned on first login
history_entries	Transcription results (full JSONB data)
presets	Saved transcription option sets
vocabulary_categories	System + user vocabulary groups
vocabularies	Word boost lists
jobs	Long-running AI job tracking

---

Key Files

File	Purpose
frontend/src/routes/+page.svelte	Main queue UI, file processing orchestration
frontend/src/lib/services/transcription.ts	AssemblyAI upload + polling with retry logic
frontend/src/lib/services/docx-export.ts	Client-side Word document generation
backend/app/api/transcription.py	Transcription submit + poll proxy
backend/app/api/ai.py	Minutes generation (Anthropic)
backend/alembic/versions/001_initial_schema.py	Database schema

---

IDE Setup

Recommended extensions:

• Svelte for VS Code
• Python
• Docker