📊 Copilot Token Usage Report2026-04-24

[github-actions]

Overview

Period: 2026-04-24T07:51Z to 2026-04-24T22:54Z (single day snapshot)
Runs analyzed: 50 (31 had token data, 19 missing token data)
Total tokens: 7,387,752 across all workflows
Effective tokens (cache-adjusted): 8,188,775
Estimated total cost: N/A — no per-token pricing available for Copilot engine
Total Actions minutes: 216 min

All token data comes from the claude-sonnet-4.6 model via the Copilot provider.

---

Workflow Summary

Workflow	Runs	Total Tokens	Avg Tokens/Run	Effective Tokens	Turns	Errors	Actions (min)
Build Test Suite	8	3,960K	495K	4,409K	10	3	43
Smoke Services	9	1,554K	173K	1,716K	14	4	46
Smoke Copilot	9	759K	84K	839K	11	3	36
Smoke Copilot BYOK	9	615K	68K	676K	14	4	34
Smoke Chroot	6	386K	64K	421K	5	2	29
Firewall Issue Dispatcher	2	114K	57K	127K	5	1	9
No token data	16	—	—	—	13	8	19

---

🔍 Optimization Opportunities

1. Build Test Suite — avg 495K tokens/run, 8 runs today

   • Highest single-workflow token consumer: ~54% of all tokens
   • Cache hit rate is excellent (~90%: 3.5M cache reads / 3.9M input tokens)
   • 3 errors across 8 runs (37.5% error rate) — investigate reliability; failed/cancelled runs still consume significant tokens before stopping
   • Consider whether all 8 runs were necessary (appears to include cancelled + retry runs)

2. Smoke Services — avg 173K tokens/run, 14 turns/run (highest turn count)

   • 14 avg turns exceeds the >15 threshold proximity — watch for growth
   • 44% error rate (4/9 runs): 2 cancelled, 1 failure, 1 failure — errors waste tokens
   • Reducing flaky runs could save ~346K tokens (2 failed runs × avg)

3. Smoke Copilot & Smoke Copilot BYOK — combined 1.37M tokens

   • BYOK variant has slightly fewer tokens but same high turn count (14/run)
   • 33–44% error rates across both — same cancelled batch causing errors
   • The cancelled batch (run IDs ending in ...525xxx) affected all smoke workflows simultaneously — likely a single CI event causing widespread cancellation

4. Workflows without token data (19 runs / 6 workflows) — these may lack api-proxy instrumentation:

   • Daily Copilot Token Usage Analyzer, Daily Claude Token Usage Analyzer
   • Issue Duplication Detector, Daily Security Review, CI/CD Gap Assessment
   • Daily Copilot Token Optimization Advisor

---

Per-Workflow Details

Build Test Suite

• Runs: 8 — §24914910413 §24910109048 §24906816580 §24904610377 §24904524970✗ §24904106794 §24903914563 §24878590443✗
• Total tokens: 3,960,339 (avg 495,042/run)
• Effective tokens: 4,409,574 (avg 551,196/run)
• Turns: 10 total (avg 1.3/run)
• Model: claude-sonnet-4.6 — input: 3,928K, output: 32K, cache reads: 3,533K (90% cache efficiency)
• Error rate: 3/8 runs (1 failure, 1 cancelled, 1 cancelled)

Smoke Services

• Runs: 9 — §24914910448 §24910109064 §24906816590 §24904610401 §24904525020✗ §24904106729 §24903914570 §24890336015✗ §24878590468✗
• Total tokens: 1,554,363 (avg 172,707/run)
• Effective tokens: 1,716,450 (avg 190,716/run)
• Turns: 14 total (avg 1.6/run)
• Model: claude-sonnet-4.6 — input: 1,546K, output: 8K, cache reads: 1,382K (89% cache efficiency)
• Error rate: 4/9 runs (2 failures, 1 cancelled)

Smoke Copilot

• Runs: 9 — §24914910425 §24910109070 §24906816547 §24904610437 §24904525028✗ §24904106742 §24903914565 §24903726121 §24889781150✗
• Total tokens: 758,538 (avg 84,282/run)
• Effective tokens: 838,717 (avg 93,190/run)
• Turns: 11 total (avg 1.2/run)
• Model: claude-sonnet-4.6 — input: 751K, output: 7K, cache reads: 587K (78% cache efficiency)
• Error rate: 3/9 runs (1 failure, 1 cancelled)

Smoke Copilot BYOK

• Runs: 9 — §24914910420 §24910109126 §24906816546 §24904610406 §24904525041✗ §24904106719 §24903914607 §24890617134✗ §24878590459✗
• Total tokens: 614,541 (avg 68,282/run)
• Effective tokens: 676,421 (avg 75,157/run)
• Turns: 14 total (avg 1.6/run)
• Model: claude-sonnet-4.6 — input: 610K, output: 5K, cache reads: 471K (77% cache efficiency)
• Error rate: 4/9 runs (2 failures, 1 cancelled)

Smoke Chroot

• Runs: 6 — §24914910410 §24910109086 §24904610414 §24904525071✗ §24904106764 §24903914593
• Total tokens: 385,898 (avg 64,316/run)
• Effective tokens: 421,101 (avg 70,183/run)
• Turns: 5 total (avg 0.8/run)
• Model: claude-sonnet-4.6 — input: 383K, output: 3K, cache reads: 261K (68% cache efficiency)
• Error rate: 2/6 runs (1 cancelled)

Firewall Issue Dispatcher

• Runs: 2 — §24905713308 §24890031096✗
• Total tokens: 114,073 (avg 57,036/run)
• Effective tokens: 126,512 (avg 63,256/run)
• Turns: 5 total (avg 2.5/run)
• Model: claude-sonnet-4.6 — input: 112K, output: 2K, cache reads: 76K (68% cache efficiency)
• Error rate: 1/2 runs (1 failure)

Workflows Without Token Data (19 runs)

These runs had token_usage = null/0 — api-proxy may not be enabled or runs exited before any LLM call:

Workflow	Runs	Errors	Action Min
Daily Copilot Token Usage Analyzer	2	1	2
Issue Duplication Detector	1	1	5
Daily Security Review and Threat Modeling	1	1	4
CI/CD Pipelines and Integration Tests Gap Assessment	1	1	4
Daily Copilot Token Optimization Advisor	1	1	2
Daily Claude Token Usage Analyzer	1	1	2

All 6 of these workflows show 1 error each — they may be consistently failing before reaching LLM calls, which would explain zero token data. Investigate whether api-proxy is configured for these workflows.

---

Historical Trend

This is the first Copilot token usage report. Future reports will compare against this baseline.

Notable observation: All smoke workflows experienced a simultaneous cancellation event (batch ...525xxx runs at ~2026-04-24T12:xx UTC), inflating error counts across the board. This single event accounts for 5 of the 23 total errors.

---

References:

• §24915516933 — This analyzer run
• §24914910413 — Build Test Suite (latest)
• §24914910448 — Smoke Services (latest)

Generated by Daily Copilot Token Usage Analyzer · ● 284.3K · ◷