mirror of
https://github.com/nightly-labs/openbot.git
synced 2026-11-15 14:01:41 +08:00
NB
ad12372347
* Make transparent pictures on the Canvas and paste pictures into it - GPT Image and Grok Imagine image steps get a Background chip. Transparent makes a picture with clear pixels. Codex makes them itself. For Grok, the step asks for a flat key colour, and the main process removes it and the colour tint at the border of the subject. - An asset or an image result with clear pixels shows with no card behind it, at its own shape. - A picture on the clipboard pastes into the Canvas as an asset at the pointer. - Joined servers: the host capability workflows-media-transparent-v1 gates the chip. Paste uses the existing upload path. * Keep the Grok picture when its clear copy fails, and fix review findings - A Grok step keeps the picture that the model made when the clear copy cannot be saved. - A clear image result is not cropped, as a clear asset is not. - The canvas keeps the last picture shape while the next one loads, so the card does not flash. - A paste adds the asset only when the canvas still takes assets when the upload ends. - The comment of clearFlatBackground says how the background grows. * Put the clear picture rule before the taller asset rule, so the CSS keeps ascending specificity