- Add _ai_prefab flag to track pre-created empty AI widget
- start_reasoning() now reorders widgets: remove empty AI widget, mount reasoning first, then remount AI widget
- This ensures reasoning always appears BEFORE the actual answer
- Filter out 'Conversation started'/'对话开始' info notifications (implicit event, no need to show)
- Increase reasoning display limit from 2000 to 3000 chars
Root cause: aibox.py think_and_respond recursively calls itself after tool
execution, and each recursive call's complete callback carries only the
final-round full_response. The old finish_ai logic treated complete as a
full-text alignment signal — when full_text != self._ai_text (which always
happened after recursive tool calls because _ai_text accumulated all chunks
while complete only held the last round), it triggered remove+remount,
causing visible duplicate/flash of AI responses. A secondary issue was
dialogue_ended arriving before complete, which made a second finish_ai call
remount the message a second time.
Changes:
- Remove the full-text replacement branch in finish_ai; chunk streaming is
now the authoritative content source, complete is a finalization signal only
- Add _ai_finalized flag: start_ai resets it, finish_ai sets it; repeated
finish_ai calls (dialogue_ended + complete) return early to prevent remount
- append_ai drops late chunks after finalization to avoid dirty appends
- Empty pre-built widget without chunks still gets filled via w.update(full_text)
- Full lost-frame fallback (no widget + full_text) is preserved
- Update BotOps/README.md protocol section to reflect new complete semantics
qing