[api-consumption] 📊 GitHub API Consumption Report — 2026-06-16

[github-actions[bot]]

📊 GitHub API Consumption Report

Report Date: 2026-06-16 · Repository: github/gh-aw · Run: #27617192820

Window analysed: 2026-06-15 12:33 UTC → 2026-06-16 12:26 UTC (rolling 24h).

---

Today at a Glance

Metric	Value
🤖 Total Runs	335 (293 ✅ / 38 ❌, 4 in-progress)
🎯 Success Rate	87.5%
🔗 GitHub REST API Calls	3,216 (core quota consumed — includes reads, writes, and all GitHub API operations)
📝 Safe-Output Writes	34 (issues + PRs + comments + discussions created by safe-output tools)
⏱ Avg Duration	601s (p95: 1,566s)

---

🔗 GitHub API Calls Trend (90 days)

Daily REST API consumption has collapsed sharply in mid-June: from the 9 Jun peak of 207,880 calls (a >2σ anomaly) down to 78,479 on 13 Jun and then a sustained low plateau of 7,411 → 2,411 → 3,216 across 14–16 Jun. The order-of-magnitude drop between 13 and 14 Jun is too abrupt to be organic and most likely reflects a change in how github_api_calls is measured by the logs pipeline rather than a true 25× efficiency gain — worth confirming before reading the absolute numbers as real quota savings.

---

🔗 GitHub API Calls by Workflow Trend (30 days)

Across the 30-day window the heaviest consumers have consistently been the high-frequency PR/CI workflows — Smoke CI, PR Sous Chef, Issue Monster, and PR Description Updater. Their per-workflow footprints have contracted in lockstep with the overall mid-June drop, which reinforces that the discontinuity is pipeline-wide (affecting every workflow uniformly) rather than isolated to any single agent.

---

🔗 GitHub REST API Calls Heatmap (90 days)

The heatmap shows consumption is fairly evenly spread across weekdays, with the most intense cells landing on the late-May / early-June weeks that contained the 27 May (118k) and 9 Jun (208k) spikes. The most recent week is uniformly cool across every weekday, consistent with the post-13 Jun plateau. No strong weekend-versus-weekday signal is visible — the scheduled agent fleet runs continuously.

---

🍩 Top API Burners (24h)

In the last 24h, API consumption is broadly distributed rather than concentrated: the top burner (Smoke CI, 222 calls) accounts for only ~7% of the 3,216 total, and the top 5 together hold roughly half. No single workflow dominates, so there is low concentration risk today — optimisation effort would need to target the long tail of many small consumers rather than one outlier.

---

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

The top REST API consumers are Smoke CI (222), Issue Monster (221), and PR Sous Chef (203), each well clustered around 200 calls. All workflows are far below the 15,000 calls/hour GitHub limit — peak per-run consumption today was only 33 calls. There is no immediate rate-limit pressure; the main optimisation opportunity is the high run-count of AI Moderator (46 runs for 135 calls), where batching or debouncing could trim redundant invocations.

---

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

Workflow	REST API Calls	Runs	Avg Duration
Smoke CI	222	30	208s
Issue Monster	221	14	552s
PR Sous Chef	203	17	591s
PR Description Updater	171	15	499s
AI Moderator	135	46	383s
Auto-Triage Issues	116	15	322s
Contribution Check	60	6	672s
PR Triage Agent	52	4	687s
[aw] Failure Investigator (6h)	52	4	964s
Design Decision Gate 🏗️	50	7	347s

---

Trending Indicators

• 7-day API trend: ↓ -67.4% vs. previous 7 days (driven by the mid-June measurement discontinuity)
• 30-day API trend: → insufficient data (29 of 60 days required for a full 30-day-over-30-day comparison)
• GitHub REST API call rate: ~26,174 calls/day averaged over the last 7 days (inflated by the 10 Jun / 13 Jun spikes; the last 3 days average just ~2,679/day) — comfortably within the 15,000/hour limit

---

📦 Cache Memory Status

• Location: /tmp/gh-aw/cache-memory/trending/api-consumption/history.jsonl
• Cache restored from previous run: yes (28 entries)
• Collection mode: backfill (entry_count 28 < 30 threshold)
• Logs start_date used: -90d (the logs tool's per-call timeout capped each page; the full 24h window was reassembled across paginated before_run_id continuations)
• Data points stored: 29
• Earliest entry: 2026-05-18
• Retention policy: 90 days

---

Automatically generated by the api-consumption-report workflow.

Generated by 📊 GitHub API Consumption Report Agent · 367.8 AIC · ⌖ 34.4 AIC · ⊞ 11.4K · ◷

• expires on Jun 19, 2026, 4:58 AM UTC-08:00