gptme-daily-briefing

Generic daily-briefing bundle schema and collectors for gptme agents.

This is the upstream extraction of Bob's collect-daily-briefing.py morning pipeline — see the design doc and the tracking issue.

What it provides

• schema — TypedDicts describing the bundle JSON contract that email and voice renderers consume (BriefingBundle, Bullets, Analytics, Workstream, etc.). Any agent can produce or consume this format.
• collectors — generic, agent-agnostic data gatherers:
  • collect_blockers(repo, label) — open issues by label
  • collect_active_tasks(workspace_root) — gptodo active/todo tasks
  • collect_waiting_tasks(workspace_root) — task frontmatter waiting_for
  • collect_recent_highlights(workspace_root) — recent commit subjects
  • collect_session_stats(sessions_dir) — gptme-sessions count + categories
  • collect_open_prs(repos, username) — open PRs by user across repos
  • collect_graphql_rate_limit() — GitHub GraphQL budget probe

Agent-specific bits (KPI shape, bandit state, agent persona) stay in each agent's local wrapper.

Use

from pathlib import Path
from gptme_daily_briefing.collectors import (
    collect_blockers,
    collect_active_tasks,
    collect_recent_highlights,
)

workspace = Path("/home/agent/agent")
bundle = {
    "bullets": {
        "blockers": collect_blockers("OWNER/REPO", "request-for-erik"),
        "active_tasks": collect_active_tasks(workspace),
        "recent_highlights": collect_recent_highlights(workspace),
    },
}

The agent's own wrapper composes the bundle and writes it to state/daily-briefing/YYYY-MM-DD.json (or wherever).

Bob's reference wrapper

scripts/monitoring/collect-daily-briefing.py in Bob's workspace is the reference consumer. It composes the generic collectors here with Bob-local additions (Thompson sampling bandit tops, weekly-goals.py KPI snapshot, pr-review-guide.py rich PR list).