Getting started

Quick start

Run the Voice API locally, define a schema, and parse your first request.

The service runs on macOS or Linux with Python 3.11+ and Ollama. The models need about 7 GB of disk.

1. Start the local language model

Terminal

ollama serve   # or open the Ollama app
ollama pull qwen2.5:7b

2. Start the Voice API

Terminal

cd voice-order-service
uv venv --python 3.12 .venv
uv pip install --python .venv/bin/python -r requirements.txt
.venv/bin/uvicorn app.main:app --port 8000

The first start downloads the speech model (~1.6 GB). curl localhost:8000/ready reports "status": "ready" when both models are usable.

3. Parse text against an example schema

No microphone needed — text runs the same extraction and validation:

curl

curl -X POST http://localhost:8000/api/v1/text/parse \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"text": "بدي ٢ شاورما ووجبة زنجر والتوصيل لخلدا", "schema": "restaurant_order"}'

4. Define your own schema

Open Create schema in the dashboard, add fields, save, and press Test schema to speak to it in the playground. Or POST the JSON to /api/v1/schemas — see Schemas.