v0.4.0-alpha.3
- Improved —
unblock/ai/model_preference filter now receives the user-selected model and can always override.
- Improved — AI tool continuation prompt extracted to dedicated prompt file.
- Improved — AI tool continuation history builder extracted to separate method for clarity.
- Fixed — AI continuation calls rejected by Anthropic due to missing
tool_result for client-side tool calls.
- Fixed — AI continuation calls failing with Google due to multiple consecutive user-role messages.
- Fixed — AI continuation function response IDs mismatching provider's original call IDs.
- Fixed — AI chat skill dropdown not opening when typing
/ after a newline.
- Fixed — AI chat skill selection via click not removing the
/ command text from the input.
- Fixed — AI chat skill selection removing text after the
/ command instead of preserving it.