1.4.8 - 4o-mini

1.4.8

1.4.7 Support for model 4o. Accept image nodes

GPT model 4o is supported.
Image nodes from canvas now sent as chat context.
Make sure you pick a model which supports image inputs. We don't have any check against image-supporting model, the GPT call will just error out.

Thank you @blaulan!

1.4.6

• Fixes bug that made some GPT model aliases fail to run (such as gpt-4-turbo-preview).

1.4.5

Update GPT model list with latest names.

1.4.4

• Fix ancestor ordering bug. Ancestors were being sent in reverse order.
  • Ancestors are included in context in the order shown on canvas: top-down parents to children, and left-to-right based on sibling X position.

1.4.3

• Token counting: Chat Stream will use all of the tokens available for the selected model (or up to the input token limit you specify).
  • If a note will not fit fully in the context, the end of the note will be truncated
• subpath support: 'Narrow to heading' and 'Narrow to block' will work as expected. Chat Stream will use the text shown on-canvas, instead of entire file.

1.3.3

• API URL setting for AI completions
• Instructions for using 'SYSTEM PROMPT' override in canvas. If your chat stream starts with a note beginning with that string, the proceeding contents of that initial note will be used as system prompt (ignoring the settings value).

1.3.2

• Temperature setting
• Fix auto-placement bug for negative x values

1.3.1

• Avoid focus change when user has selected a different node
• Fix typo in settings panel

1.3.0

• Use commands for proper hotkey support
  • This also fixes a bug where plugin would sometimes not respond to hotkeys