Skip to content

README.md

Latest commit

 

History

History
86 lines (61 loc) · 2.57 KB

README.md

File metadata and controls

86 lines (61 loc) · 2.57 KB

Echo

Echo is a text-to-speech workflow app that records audio, transcribes it to text, then uses AI to polish and structure the text. It is designed for fast capture, review, and publishing-ready outputs.

What you can create

  • Blog post: transcribe, review, and refine into a publishable post
  • Interview: dialog format that preserves each speaker's personality
  • Notes: quick thought capture and clean transcription
  • News brief: factual summary with clear points and bullets

How it works

  1. Record or import audio
  2. Transcribe audio to text
  3. Review and edit the transcript
  4. Choose an output format: Blog, Interview, Notes, or News
  5. Let AI refine the text to match the selected structure
  6. Copy or export the final result

Architecture

Echo system architecture

Echo currently runs as a backend workflow made of an API service, a background worker, shared SQLite persistence, S3-compatible object storage, and a RabbitMQ broker:

apps/api/       .NET Minimal API for recording ingestion and metadata
apps/worker/    Go worker for queued transcription jobs
SQLite          shared database for recordings and transcription jobs
MinIO/S3        object storage for uploaded audio files
RabbitMQ        message broker for asynchronous workflows

Run with Docker Compose

From the project root, create a local environment file:

cp .env.example .env

Fill in the required values in .env:

MINIO_ROOT_USER=minioadmin
MINIO_ROOT_PASSWORD=minioadmin
TRANSCRIPTION_PROVIDER=openrouter
OPENROUTER_API_KEY=your-openrouter-api-key
OPENROUTER_TRANSCRIPTION_MODEL=openai/gpt-4o-mini-transcribe

Start the local stack:

docker compose up --build

The Compose stack starts:

  • api: .NET API on http://localhost:8080
  • worker: Go background worker for transcription jobs
  • minio: S3-compatible object storage on http://localhost:9000
  • minio console: http://localhost:9001
  • rabbitmq: RabbitMQ broker on localhost:5672
  • rabbitmq management UI: http://localhost:15672

The API and worker share a SQLite database through the appdata Docker volume. Uploaded audio files are stored in MinIO in the echo-files bucket, which is created automatically by the minio-init service.

RabbitMQ uses the default local credentials echo / echo in Compose.

Useful commands:

docker compose logs -f api worker rabbitmq
docker compose down
docker compose down -v

Use docker compose down -v only when you want to remove the SQLite and MinIO Docker volumes.

License

MIT - see LICENSE.