💡 Proactive Workspace Monitoring: Avatar-Initiated Project Status Briefings

[github-actions[bot]]

Summary

The avatar proactively monitors workspace state (git changes, new issues, PR reviews needed, file modifications) and surfaces relevant updates when the user opens TalkTerm or at user-configured intervals. Unlike one-time workspace scanning, this provides ongoing situational awareness — the avatar becomes a colleague who stays current on your project.

Market Signal

GitHub Copilot has workspace awareness. Codex monitors projects with always-on presence. Manus offers remote monitoring from any device. However, none use an avatar or voice to deliver updates — they all rely on text notifications. Gartner predicts 40% of enterprise apps will integrate task-specific AI agents by end of 2026. The ambient computing trend points toward proactive, context-aware assistants rather than reactive query-response tools.

User Signal

TalkTerm's PRD describes the avatar greeting returning users by name and offering to continue recent sessions (FR33). The Contextual Workspace Intelligence idea (#211) proposed initial project scanning on launch. This proposal extends that concept to ongoing, periodic monitoring — the difference between "scan once" and "stay aware." The PM persona (Sarah) would benefit enormously from "Mary tells you about 3 new comments on your PRD before you even ask."

Technical Opportunity

Local workspace monitoring (git status, file watchers) is zero-cost and zero-latency. The IPC bridge (FR15) already supports main-to-renderer event streaming. System tray integration enables background presence. Light status checks via Claude Haiku minimize API costs. The DomainEventBus pattern enables decoupled monitoring modules. Session persistence (FR32) stores monitoring state across launches.

Assessment

Dimension	Score	Rationale
Feasibility	med	Local monitoring is trivial; API-based checks add cost considerations
Impact	high	Transforms the avatar from reactive to proactive — the "colleague who stays current"
Urgency	med	Competitors are adding ambient awareness; TalkTerm's avatar voice makes this uniquely powerful

Adversarial Review

Strongest objection: Proactive behavior risks being annoying. If the briefing isn't relevant, it becomes noise that erodes the trust relationship with the avatar — the exact opposite of TalkTerm's "companion, not tool" design principle.

Rebuttal: Opt-in by design. The avatar asks "Would you like me to keep an eye on your project?" and lets the user set scope and frequency. Monitoring starts with free/local signals (git status, file changes, local issue cache) before any API calls. The preference memory (FR51) learns which updates the user acts on vs. dismisses, refining relevance over time. Cost is managed by using Haiku for status checks and batching queries. The value proposition for the PM persona is compelling and directly supports the "colleague who remembers" experience principle.

Suggested Next Step

Define the monitoring signal taxonomy (git events, file changes, issue tracker polling, calendar). Design the opt-in configuration flow and briefing presentation UX. Prototype local-first monitoring (git status + file watchers) with no API cost.