- Install Ollama and start it on this device. Choose a chat model from the Ollama model library, then follow its download instructions.
- Keep Ollama on
127.0.0.1:11434. For local-only runtime settings, set OLLAMA_NO_CLOUD=1 in Ollama's environment and restart it.
- For a hosted page, set
OLLAMA_ORIGINS to the exact page origin, restart Ollama, and allow loopback access if your browser asks.
If you're in an embedded browser, open this page in a browser tab. If your browser blocks a hosted page from reaching Ollama, serve the downloaded site at http://127.0.0.1:8080/ai.html. From the site's dist directory, run python -m http.server 8080 --bind 127.0.0.1.
On a phone, this console needs an Ollama-compatible runtime on the phone itself. The site layout works on mobile, and the loopback address always refers to the device you're using.
Ollama setup docs ↗