fix(ai): report compacted model context - #312

Summary

• feed the chat context pill from the existing native chatBudget endpoint, which measures the messages actually sent to the model
• retain the renderer transcript estimate as an immediate/browser/error fallback
• refresh usage whenever the active session's visible message array changes, including after manual compaction
• ignore late budget responses from a previously selected session

This is a follow-up to #238: preserving the complete transcript was correct, but it meant the transcript-based meter no longer reflected the compacted model payload.

Related issue

Related to #238

Type of change

• Bug fix
• Feature
• Enhancement
• Documentation
• Refactor / maintenance
• Performance
• Security

Release impact

• Patch
• Minor
• Major / breaking change
• No release note needed

Desktop impact

• Windows
• macOS
• Linux
• Installer / packaging
• Not platform-specific

Screenshots / video

Not included; the context pill retains its existing UI and now receives the correct post-compaction value.

Testing

• npx vitest --run src/components/ai-edition/useChatBudget.test.ts (3 passed)
  • renderer estimate before native usage arrives
  • stale response ignored after session switch
  • unchanged visible transcript refreshes from 500 to 40 model tokens after compaction
• npm run test (1,680 passed, 1 skipped across 141 files)
• npx tsc --noEmit
• npx tsc -p tsconfig.test.json --noEmit
• npx biome check on all four changed files
• npm run docs:check
• npm run i18n:check
• npm run build-vite

Authored with Codex assistance and manually verified against the native modelMessages(session) budget path.