💡 Asynchronous Task Review Queue: Avatar-Narrated Background Results Briefing

[github-actions[bot]]

Summary

When TalkTerm completes background or scheduled tasks, results land in a review queue rather than disappearing silently. On next launch or return, the avatar greets the user with a conversational briefing: "While you were away, I finished 3 tasks. The competitive analysis flagged something interesting — want to start there?" Each queued result includes a summary card, full output, and approve/revise/discard options presented as ActionCards.

Market Signal

OpenAI Codex Automations (2026) established the pattern: background tasks run on schedules, results land in a review queue for user action. Claude Dispatch delivers completed task results to the user's phone. The 2026 agentic AI trend report highlights that autonomous task completion without a review loop creates trust and accountability problems. 87% of voice AI users still want human oversight (G2 2026), confirming that "fire and forget" is insufficient for trust-building.

User Signal

Existing idea #232 (Goal Mode) covers checkpoint-based background workflow execution with avatar notifications. However, #232 focuses on the execution side (running tasks in background with checkpoints). This proposal addresses the completion side: what happens when the user returns? The review queue is the accountability layer that closes the loop. Without it, background task results are notifications to be dismissed rather than structured outputs to be reviewed and acted upon.

Technical Opportunity

Session persistence (Epic 8, SQLite) and audit trail (Epic 9) already provide the storage infrastructure for queued results. The review UI uses existing ActionCards (UX-DR2) for approve/revise/discard and OutputPanel (UX-DR4) for full result display. The avatar greeting flow (FR33) already offers to "continue the most recent incomplete session" — extending this to "review completed background tasks" is a natural evolution. The queue is a lightweight layer on existing session storage, not a new system.

Assessment

Dimension	Score	Rationale
Feasibility	high	Built on existing session persistence (Epic 8) and audit trail (Epic 9); review UI uses existing ActionCard and OutputPanel components
Impact	high	Closes the background execution trust loop; makes autonomous task completion accountable and actionable
Urgency	med	Should be designed alongside Goal Mode (#232) to ensure clean separation; not time-sensitive to external events

Adversarial Review

Strongest objection: This overlaps significantly with Goal Mode (#232), which already includes "avatar notifications" for background task completion.

Rebuttal: Notifications and review queues serve fundamentally different purposes. A notification says "done" — it's informational and dismissible. A review queue says "here are your results, let's walk through them together" — it's interactive and accountability-building. The review queue includes: (1) avatar-narrated priority ordering ("the analysis flagged something interesting"), (2) structured approve/revise/discard actions per result, (3) batch operations ("approve all routine tasks"), and (4) the ability to revise and re-run tasks that didn't meet expectations. This is the completion UX that makes background execution trustworthy for non-technical users.

Suggested Next Step

Define the review queue data model as an extension to the session persistence schema (Epic 8). Design the avatar briefing flow: on launch, check for completed background tasks → avatar summarizes queue → user navigates results via ActionCards → approve/revise/discard each result. Specify the relationship between this feature and Goal Mode (#232) to ensure clean separation of concerns.