[api-consumption] 📊 GitHub API Consumption Report — 2026-07-12

[github-actions[bot]]

📊 GitHub API Consumption Report

Report Date: 2026-07-12 · Repository: github/gh-aw · Run: #29190160964

⚠️ Partial dataset for today. The logs collector timed out after 120s mid-download, so only 99 of ~184 typical daily runs had complete artifacts on disk (183 run directories were empty). Today's absolute totals (API calls, safe-output writes, run counts) therefore understate actual 24h activity and today's point is annotated as partial on the trend charts. Trend percentages and the call-rate figure below are computed from complete historical days only and are unaffected.

---

Today at a Glance

Metric	Value
🤖 Total Runs	99 (89 ✅ / 10 ❌) (partial)
🎯 Success Rate	89.9%
🔗 GitHub REST API Calls	16,152 (partial — core quota consumed; includes reads, writes, and all GitHub API operations)
📝 Safe-Output Writes	165 (partial — issues + PRs + comments + discussions created by safe-output tools)
⏱ Avg Duration	611s (p95: 1,053s)

---

🔗 GitHub API Calls Trend (90 days)

Over the full history, daily REST API consumption on complete days has averaged roughly 104k calls/day across the last 7 days, up sharply from the prior week. Two spikes dominate the window: Jun 09 (~208k) and a standout Jun 18 (~573k), both well above the 30-day rolling average (purple dashed) which sits near 60–65k. The amber point marks today's partial collection (16.2k) and should not be read as a genuine drop.

---

🔗 GitHub API Calls by Workflow Trend (30 days)

The top-6 consumers over the last 30 days show that a handful of workflows drive most of the quota, with sporadic high-consumption days rather than steady baselines — consistent with schedule-triggered batch runs. No single workflow sustains a runaway upward slope; peaks are episodic and correlate with the whole-fleet spike days seen in the 90-day chart.

---

🔗 GitHub REST API Calls Heatmap (90 days)

The week × weekday view highlights that consumption is concentrated in a few hot cells rather than evenly spread, with the mid-June week carrying the darkest band (the ~573k Jun 18 outlier). Weekends are generally lighter than weekdays, reflecting development-driven PR and triage workflows tapering off outside the work week.

---

🍩 Top API Burners (24h)

Within the (partial) 24h sample, jsweep — JavaScript Unbloater alone consumed ~4.9k calls in a single run (~31% of the captured total), and PR Sous Chef added ~2.5k across 12 runs. The top three workflows account for roughly half of all captured consumption, indicating meaningful concentration worth watching if these scale up.

---

🔗 GitHub REST API Consumption by Workflow (last 24h)

The per-workflow bars confirm jsweep and PR Sous Chef as the leading REST consumers in the captured window. Even the top burner (~4.9k) sits comfortably under the 15,000/hr core limit, so there is no immediate throttling risk; the optimisation opportunity is jsweep's high cost-per-run (~4.9k in one run) versus PR Sous Chef's lower ~204 calls/run amortised over 12 runs.

---

Top 10 Workflows by REST API Consumption (last 24h)

Workflow	REST API Calls	Runs	Avg Duration
jsweep — JavaScript Unbloater	4,945	1	1,090s
PR Sous Chef	2,450	12	495s
Smoke CI	1,101	4	233s
PR Description Updater	986	3	453s
PR Triage Agent	903	2	838s
Auto-Triage Issues	813	3	420s
AI Moderator	654	1	1,028s
Impeccable Skills Reviewer	602	2	427s
PR Code Quality Reviewer	596	2	434s
Smoke Copilot	591	3	834s

---

Trending Indicators

• 7-day API trend: ↑ +158.8% vs. previous 7 days (complete days only; recent week ~104k/day vs. prior week driven higher by the Jun 18 spike)
• 30-day API trend: → (insufficient data — 53 days of history; needs 60 for a 30-vs-30 comparison)
• GitHub REST API call rate: ~104,204 calls/day over last 7 complete days (hourly limit: 15,000)

❌ Failed runs in captured window (10)

Workflow	Conclusion	Count
Smoke CI	failure	4
PR Triage Agent	failure	2
Smoke OpenCode	failure	1
Documentation Noob Tester	failure	1
GitHub Remote MCP Authentication Test	failure	1
daily-experiment-report	failure	1

Smoke CI accounts for 4 of 10 failures in the captured sample and is worth a closer look.

---

📦 Cache Memory Status

• Location: /tmp/gh-aw/cache-memory/trending/api-consumption/history.jsonl
• Cache restored from previous run: yes (52 entries)
• Collection mode: incremental
• Logs start_date used: -1d
• Data points stored: 53
• Earliest entry: 2026-05-18
• Retention policy: 90 days

i️ Collection notes

• The agenticworkflows logs call for start_date=-1d returned HTTP timeout after 120s; 99 run directories had fully downloaded artifacts and were used as the authoritative dataset per the collection policy. No further continuation/retry was attempted because a substantial dataset was already on disk and a retry would likely re-hit the same timeout.
• All metrics were parsed in-memory from run_summary.json (github_rate_limit_usage.core_consumed), aw_info.json, and summary.json (safe_outputs.total_items).

---

Automatically generated by the api-consumption-report workflow.

Warning

Firewall blocked 1 domain

The following domain was blocked by the firewall during workflow execution:

• awmgmcpg

To allow these domains, add them to the network.allowed list in your workflow frontmatter:

network:
  allowed:
    - defaults
    - "awmgmcpg"

See Network Configuration for more information.

Generated by 📊 GitHub API Consumption Report Agent · 225.9 AIC · ⌖ 41.9 AIC · ⊞ 11.9K · ◷

• expires on Jul 15, 2026, 3:22 AM UTC-08:00

[github-actions[bot]]

This discussion has been marked as outdated by GitHub API Consumption Report Agent.

A newer discussion is available at Discussion #45255.