Use case
Offline and private AI coding
The private path
Install Ollama, pull a coder model, and use the desktop app or the CLI. The privacy policy states that the desktop app stores its model config on your computer and does not send a project to codesyncsi.com unless you upload it.
A hosted run is different. Files go to Cloudflare R2, and a model call can send a clipped excerpt to Workers AI or to the provider whose key you stored. Do not upload a repository you are not allowed to process.
When the network is down
Local scan, diagnose, and safe repair do not need a model. Generate tries the configured model and falls back to the labeled template if Ollama is not answering. That fallback is intentional so the command still writes a file you can see.