(CODUX-191) Preserve chat focus after widget send

[numnx]

🧩 Task: Preserve chat focus after widget send

Automated task PR opened by Code UX for T06.

📋 Summary

Sprint: Sprint 191 · Persist chat input history · Task: T06

Store each user's latest chat input in the database so they can leave the page and return with the last message still present in all chat inputs, and navigate recent messages with ArrowUp and ArrowDown.

🤖 Model & Provider

Provider	Codex CLI
Model	gpt-5.5

⏱️ Timing

Started	2026-07-08 21:50:32 UTC
Finished	2026-07-08 21:53:44 UTC
Duration	3m 12s

📊 Token Usage

Metric	Value
Input tokens	116,190
Cached input tokens	1,232,384
Output tokens	5,221
Total tokens	1,353,795
Invocations	1
Tool calls	37
Active time	2m 58s

Included usage estimate (subscription/local login): $1.35

✅ QA Review

⏳ QA review runs after this PR is opened. Check the task card in the dashboard for the outcome once it lands.

📝 Full Task Prompt

Objective: Keep chat focus and draft flow predictable after a rich-widget suggestion sends immediately.

Scope:
- Ensure widget-triggered sends leave the composer in a usable state.
- Preserve the draft-restoration flow after the send.

Requirements:
- Focus remains sensible for continued chatting after a suggestion sends.
- No stale text is left behind.
- Widget-triggered sends do not interfere with draft persistence.

Constraints:
- Keep the behavior consistent across desktop and browser surfaces.
- Avoid regressions in keyboard navigation.

Verification:
- Manual interaction confirms the focus and draft state are correct after widget sends.
- UI tests cover the focus path.

🌿 Branch Info

Base: feature/CODUX-191-persist-chat-input-history
Head: task/feature-codux-191-persist-chat-input-history-t06-codex-mrcm3bpx

---

🤖 Generated by Code UX