fix: avoid Tauri resource aborts in Auto Suggest

[cubicj]

PR Checklist

• Required Checks
  • Have you added type definitions?
  • Have you tested your changes?
  • Have you checked that it won't break any existing features?
• If your PR uses models¹, check the following:
  • Have you checked if it works normally in all models?
  • Have you checked if it works normally in all web, local, and node-hosted versions? If it doesn't, have you blocked it in those versions?
• If your PR is highly AI generated², check the following:
  • Have you understood what the code does?
  • Have you cleaned up any unnecessary or redundant code?
  • Is it not a huge change?
    • We currently do not accept highly AI generated PRs that are large changes.

Summary

Fixes Auto Suggest triggering Tauri HTTP resource errors such as The resource id ... is invalid.

Auto Suggest used an AbortController to cancel in-flight suggestion requests when chat state changed, the chat page changed, or suggestions were re-rolled. On Tauri desktop, that abort signal is passed into @tauri-apps/plugin-http, where aborting around response body reads can race with native resource cleanup and surface an invalid resource id error.

Related Issues

Fixes #1443.

Changes

src/lib/ChatScreens/Suggestion.svelte

• Replace Auto Suggest's request cancellation state with a request id guard
• Avoid passing a native abort signal for Auto Suggest requests on Tauri
• Keep abort behavior for non-Tauri environments
• Only write suggestion results when the request id, selected character, and chat page still match
• Stop using doingChat.set(true); doingChat.set(false) to re-roll suggestions
• Invalidate pending suggestion results when the component is destroyed

src/ts/globalApi.svelte.ts

• Return a normalized failed fetch result when Tauri HTTP fetch throws

Impact

• Auto Suggest no longer uses Tauri native HTTP aborts, avoiding the plugin-http resource id race.
• Stale Auto Suggest responses are discarded instead of being written to the current chat.
• Re-rolling suggestions no longer toggles the global chat-in-progress state.
• On Tauri, canceled Auto Suggest requests may still finish on the backend, but their results are ignored if they are no longer current.
• Existing stored data and database schema are unchanged.
• Web and non-Tauri abort behavior is preserved.

Additional Notes

Validated with:

• pnpm check
• pnpm test
• pnpm build

git diff --check only reported existing CRLF conversion warnings.

Footnotes

1. Modifies the behavior of prompting, requesting, or handling responses from AI models. ↩

2. Over 80% of the code is AI generated. ↩