Try the API without hardware

When to use this

The mock engine returns deterministic test output instead of a model’s answer. Use it to try the API on a machine without Intel hardware (including macOS), to develop against Cascadia, or to test it in CI. It covers the complete API path, so the same requests work once you switch to a real model.

To serve a real model instead, follow the quickstart.

1. Get Cascadia

On Linux or Windows, download a release archive and use ./cascadia or .\cascadia.exe from the unpacked folder in the steps below.

On macOS, or to build from source, you need Git and Rust 1.89 or later:

git clone https://github.com/labscommunity/cascadia.git
cd cascadia
cargo build --release -p cascadia

The binary is ./target/release/cascadia. This stub build has no OpenVINO, so only the mock engine works. See installation for details.

2. Start the mock server

./target/release/cascadia run mock-model --engine mock --api 127.0.0.1:8000

Keep this terminal open. The API is now listening on port 8000.

3. Send a request

In a second terminal:

curl http://127.0.0.1:8000/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "mock-model",
    "messages": [{"role": "user", "content": "Hello, Cascadia!"}]
  }'

A JSON chat-completion response confirms the API and engine are working together. The mock engine echoes words; it does not generate a real answer.

What comes next?

• Connect an existing application to the API.
• Serve a real model on Intel hardware.