Support
Help with Canis.
Send support notes, TestFlight feedback, App Store review questions, or technical issues through the form below. Canis is local-first, so please include device model, iOS version, app version, and whether Sources were enabled when reporting a bug.
Send a message
Use this form for app support, model download issues, source-document bugs, TestFlight notes, or enterprise questions about the disposition readout.
What to know before testing.
Do I need an account?
No. Canis runs without login. The app is designed as a local-first assistant rather than an account-based cloud service.
Why does the model download take time?
Canis runs the language model on your device, so the model has to be downloaded on first use. Use Wi-Fi, keep the app open, and expect the initial setup to take longer than a normal app download.
Which model does the iOS app use?
The iOS app uses Apertus v1.1 4B with MLX for local inference. The Docker hackathon version uses Apertus v1.5 8B. The same disposition-readout idea applies across both open-weight models.
What are Sources?
Sources are documents you add so Canis can answer from your own material. The app chunks and indexes supported files locally, retrieves relevant passages, and shows citations when a response uses them.
Do my documents leave my phone?
Canis is designed for local document use. Sources, chunks, local indexes, prompts, and chat history stay on the device unless you explicitly choose to share feedback or diagnostic details.
Does the dog show emotions?
No. The dog is a readable instrument panel for model behavior. Confident, unsure, curious, concerned, reluctant, and warm describe measured response signals, not feelings or intent.
Is the disposition readout the same as answer correctness?
No. A confident answer can still be wrong. The readout is an additional transparency signal that helps you decide when to verify, ask follow-up questions, or inspect cited Sources.
Why is the iOS app faster than the Docker demo?
The iOS app is optimized for Apple Silicon through MLX and uses a smartphone-class model. The Docker version prioritizes reproducibility for judges and can be much slower on CPU-only hardware.
What should I include in a bug report?
Please include the device model, iOS version, app version, whether Sources were enabled, the kind of file involved if relevant, and the shortest steps that reproduce the issue.
Can I use Canis for medical, legal, or financial advice?
Use Canis as an experimental information tool. For high-stakes decisions, verify answers with authoritative sources or qualified professionals.