fix: ensure AI Assistant restart clears conversation history (#3261)

[pierreeurope]

Fixes #3261

This PR fixes a bug where clicking "Restart Chat" in the AI Assistant would clear the UI but continue sending the entire previous conversation history to LLM providers in subsequent API calls.

Problem:
The component was remounting (via the key prop) before the Redux state update had fully propagated, causing initialMessages to still contain the old conversation when the component re-initialized.

Solution:
Wrap the setRestartKey call in a setTimeout(0) to ensure it executes after the Redux state update has been applied. This guarantees that when the component remounts, aiAssistant?.messages is already empty.

Impact:

• Eliminates wasted API tokens (potentially hundreds or thousands per conversation)
• Reduces API costs
• Improves response times
• Ensures clean restart behavior

The fix is minimal and preserves all existing functionality while correctly resetting the conversation state.

[Copilot]

Pull request overview

Fixes an AI Assistant “Restart Chat” bug where the UI resets but subsequent LLM requests still include the prior conversation history, by ensuring both Redux and the assistant library state are cleared before forcing a remount.

Changes:

• Clear Redux aiAssistant.messages when restarting chat.
• Invoke the assistant library’s restartChat to reset internal state.
• Defer the remount trigger (restartKey increment) to the next tick to avoid remounting with stale initialMessages.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

setTimeout schedules a state update that can still fire after AiAssistantComponent unmounts (e.g., when switching to config view or closing the panel), which can trigger React warnings and unnecessary work. Consider storing the timeout id in a ref and clearing it in a useEffect cleanup (or guarding the callback with an isMounted ref) before calling setRestartKey.

[lixun910]

@copilot check if this comment makes sense and if yes, apply changes based on this feedback