Gemma 4 made local LLMs feel practical, private, and finally useful on everyday hardware.

Local → Remote Files are streamed as chunked tar batches over SSH. Remote runs tar xf - to extract on the fly Remote → Local Remote runs tar cf -, local extracts with streaming extraction — files ...

* Default idle timeout for LLM streaming responses in milliseconds. * If no token is received within this time, the request is aborted. * Set to 0 to disable (never timeout). * Default: 60 seconds.