compression▌

When agent sessions generate millions of tokens of conversation history, compression becomes mandatory. The naive approach is aggressive compression to minimize tokens per request. The correct optimization target is tokens per task: total tokens consumed to complete a task, including re-fetching costs when compression loses critical information.

Compress long AI conversations to fit context windows while preserving critical information.

Compress screenshot images (PNG/JPEG) in place while keeping the original format. Uses pngquant for PNG and jpegoptim for JPEG — both are highly effective for screenshot content (UI elements, text, flat colors).