[api-consumption] 📊 GitHub API Consumption Report — 2026-06-23

[github-actions[bot]]

📊 GitHub API Consumption Report

Report Date: 2026-06-23 · Repository: github/gh-aw · Run: #28023584929

⚠️ Partial window: log collection covered ~06:14–11:37 UTC (continuation paging timed out). Today's totals reflect this ~5.5h slice, not a full 24h, so absolute counts are undercounted versus full-day history entries.

---

Today at a Glance

Metric	Value
🤖 Total Runs	49 (46 ✅ / 3 ❌)
🎯 Success Rate	93.9%
🔗 GitHub REST API Calls	1,825 (core quota consumed — reads, writes, and all GitHub API operations; all 49 runs reported core_consumed)
📝 Safe-Output Writes	unavailable in this batch (no safe_output records surfaced)
⏱ Avg Duration	714.7s (p95: 1746.0s)

---

🔗 GitHub API Calls Trend (90 days)

Daily core-quota consumption remains highly volatile across the 35-day history, driven mostly by how much of each day's window the collector captures before timing out. The 7-day rolling average is trending up (+56.6% vs. the prior 7 days), and one day registers as a >2σ anomaly. Today's point (1,825) sits low because the window is partial.

---

🔗 GitHub API Calls by Workflow Trend (30 days)

A small set of write-heavy workflows — PR Sous Chef, the Smoke jobs, Issue Monster, and the various PR reviewers — repeatedly dominate per-day quota, while the long tail of daily/maintenance workflows each stays in single digits. The leaderboard reshuffles day to day, but the top consumers are consistently PR- and issue-mutating agents rather than read-only reporters.

---

🔗 GitHub REST API Calls Heatmap (90 days)

Consumption concentrates on weekdays, with the heaviest cells mid-week when PR and review activity peaks; weekends are markedly lighter. The brightest cell corresponds to the flagged anomaly day. Per-cell totals are biased downward on days where the collector window was partial.

---

🍩 Top API Burners (24h)

PR Sous Chef alone accounts for ~30% of the window's 1,825 core quota (546 across 4 runs), and Smoke Claude on Copilot adds another ~17%. The top three workflows together consume over half of all quota, so consumption is concentrated rather than broadly distributed — a handful of mutating agents drive the footprint.

---

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

The top consumer (PR Sous Chef, 546) is well under the 15,000/hr core limit, and no single workflow approaches the ceiling within this window. The main optimisation opportunity is PR Sous Chef's ~137 core/run average — worth checking whether its PR reads can be batched or cached — while the dozens of single-call maintenance workflows are already negligible.

---

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

Workflow	REST API Calls	Runs	Avg Duration
PR Sous Chef	546	4	9.1m
Smoke Claude on Copilot	315	2	4.2m
Issue Monster	155	4	6.1m
Smoke CI	135	1	5.8m
Contribution Check	112	1	10.6m
Auto-Triage Issues	91	1	5.8m
Go Fan	71	1	9.7m
PR Code Quality Reviewer	58	1	12.7m
Schema Consistency Checker	57	1	5.6m
Matt Pocock Skills Reviewer	28	1	9.7m

---

Trending Indicators

• 7-day API trend: ↑ +56.6% vs. previous 7 days
• 30-day API trend: → (insufficient data — 35 of 60 required days)
• GitHub REST API call rate: ~86,772 core quota/day averaged over the last 7 days (hourly core limit: 15,000 → ~360,000/day theoretical ceiling; well within budget)

---

📦 Cache Memory Status

• Location: /tmp/gh-aw/cache-memory/trending/api-consumption/history.jsonl
• Cache restored from previous run: yes (34 entries)
• Collection mode: incremental
• Logs start_date used: -1d
• Data points stored: 35
• Earliest entry: 2026-05-18
• Retention policy: 90 days

⚠️ Data Limitations

• Partial window: the initial logs(start_date="-1d") call returned 49 runs (06:14–11:37 UTC) with a continuation token; the single allowed continuation call timed out, so collection stopped per policy and today's totals cover only that ~5.5h slice.
• API source: github_api_calls here is github_rate_limit_usage.core_consumed read per-run from run_summary.json (response-header delta). The logs summary's github_api_calls field is a different, capped metric and was not used.
• Safe-output writes: no safe_output records were present in this batch, so safe-output write counts are reported as unavailable rather than zero.

---

Automatically generated by the api-consumption-report workflow.

Generated by 📊 GitHub API Consumption Report Agent · 207.2 AIC · ⌖ 41.2 AIC · ⊞ 11.4K · ◷

• expires on Jun 26, 2026, 3:57 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 #41214.