Skip to content

See Yagami in action

Yagami routes requests to local or cloud models while keeping detected private information on your machine under the default policy.

Open or download the demo (MP4, 59 seconds). The recording is silent; captions explain the routing decisions.

What the recording shows

  • A natural-language question about sunlight, answered by a local model.
  • A fictional patient chart-review request classified as sensitive and forced to local inference. The routing panel shows the decision.
  • A new conversation requesting an image. Yagami selects the configured Stability AI provider and shows the cloud route.

The final screen shows how to install Yagami:

pip install yagami

Run it yourself

For a quick local introduction, run yagami demo. It uses your configured Ollama model if available, otherwise a clearly labeled policy-only fallback. This mode deliberately disables cloud routing.

To use local and cloud models as shown in the video, run yagami init, configure your local model and cloud providers, then start yagami serve. Cloud providers require their own credentials and may charge for usage. See configuration and deployment.

About this recording

Recorded with the published Yagami 0.8.5 package, Gemma 3 via Ollama for local answers, and Stability AI for image generation. The patient information is synthetic. Responses and routing decisions are real; captions and animated yellow highlights were added for presentation.

The video demonstrates the configured policy; automated detection can miss sensitive information. Use caller-declared sensitivity or local-only policies for strict workflows. Local inference alone does not establish HIPAA compliance. See the threat model.