Skip to content

📊 Copilot Token Usage Report2026-04-02 #1591

@github-actions

Description

@github-actions

Overview

Period: 2026-04-01T01:01Z to 2026-04-02T01:01Z (24 hours)
Runs analyzed: 5 Copilot-engine runs (3 had token data)
Total tokens: 642,938 across all instrumented workflows
Estimated total cost: $1.64

⚠️ This is the first token usage report. Historical trend data will be available in future reports.

All three instrumented workflows show 0% prompt cache hit rate — a systemic optimization gap. All traffic routes through the Copilot provider to claude-sonnet-4.6.

Workflow Summary

Workflow Run Requests Total Tokens Billable Tokens Est. Cost Cache Rate I/O Ratio Model
Smoke Copilot §23878258933 10 428K 428K $1.09 0% 201:1 sonnet-4.6
Build Test Suite §23877180905 4 145K 145K $0.38 0% 79:1 sonnet-4.6
Secret Digger (Copilot) §23877933272 2 70K 70K $0.18 0% 179:1 sonnet-4.6

🔍 Optimization Opportunities

  1. All workflows — 0% cache hit rate (systemic)

    • No cache_read_tokens or cache_write_tokens in any request across all 3 workflows
    • Prompt caching is not enabled on the Copilot inference endpoint or the API proxy is not setting cache breakpoints
    • Recommendation: Enable Anthropic prompt caching on the system prompt / static context. For Smoke Copilot with 40–44K input tokens per request, caching the static prefix could reduce input cost by ~90%, saving ~$0.90/run
  2. Smoke Copilot — high cost ($1.09/run), very high I/O ratio (201:1)

    • 10 requests with input tokens growing from 40K → 44K (accumulating conversation context without caching)
    • 201:1 input/output ratio indicates large system prompt / tool schema relative to generated output
    • Recommendation: Review MCP tool surface exposed to the agent — fewer tools means smaller context. Each request reprocesses 40K+ tokens from scratch.
  3. All workflows — I/O ratios well above 10:1

    • Secret Digger (Copilot): 179:1 (34–35K input, minimal output)
    • Build Test Suite: 79:1 (34–38K input per request)
    • These ratios suggest large static context (system prompts, tool schemas) being resent every request
    • Recommendation: Identify and cache the static prefix in system prompts to reduce billable input tokens
Per-Workflow Details

Smoke Copilot

  • Run: §23878258933 (scheduled, main)
  • Requests: 10 (all claude-sonnet-4.6 via copilot provider)
  • Tokens: 428K total — 426K input, 2K output, 0 cache_read, 0 cache_write
  • Cache hit rate: 0%
  • Avg latency: 3,992ms/request
  • Estimated cost: $1.09
  • Pattern: Input tokens grow steadily (40K → 44K) across 10 requests — conversation history accumulating without caching; output is small (24–354 tokens each)
# Timestamp Input Output Duration
1 00:56:02Z 40,272 354 5,551ms
2 00:56:08Z 41,344 256 4,101ms
3 00:56:13Z 41,658 219 3,585ms
4 00:56:17Z 42,060 263 3,420ms
5 00:56:22Z 42,377 262 4,380ms
6 00:56:27Z 43,073 172 3,418ms
7 00:56:32Z 43,312 259 4,555ms
8 00:56:39Z 43,638 240 6,293ms
9 00:56:42Z 43,895 65 2,426ms
10 00:56:44Z 43,977 24 2,195ms

Secret Digger (Copilot)

  • Run: §23877933272 (scheduled, main)
  • Requests: 2 (all claude-sonnet-4.6 via copilot provider)
  • Tokens: 70K total — 70K input, 391 output, 0 cache_read, 0 cache_write
  • Cache hit rate: 0%
  • Avg latency: 5,277ms/request
  • Estimated cost: $0.18
# Timestamp Input Output Duration
1 00:44:31Z 34,706 340 7,808ms
2 00:44:34Z 35,118 51 2,746ms

Build Test Suite

  • Run: §23877180905 (PR fix/token-analyzer-domains-and-scope)
  • Requests: 4 (all claude-sonnet-4.6 via copilot provider)
  • Tokens: 145K total — 143K input, 1.8K output, 0 cache_read, 0 cache_write
  • Cache hit rate: 0%
  • Avg latency: 5,995ms/request
  • Estimated cost: $0.38
# Timestamp Input Output Duration
1 00:17:27Z 34,019 345 5,697ms
2 00:17:39Z 34,570 1,104 9,239ms
3 00:17:45Z 38,064 209 5,207ms
4 00:17:49Z 36,543 149 3,837ms
Workflows Without Token Data

The following Copilot-engine workflows ran in the past 24 hours but had no agent-artifacts with token data:

Workflow Run ID Reason
Agentic Maintenance §23878421596 No agent-artifacts artifact
Smoke Services §23877180900 No agent-artifacts artifact (has agent artifact but no token logs)
Firewall Issue Dispatcher §23878309621 No agent-artifacts artifact

These workflows may not use --enable-api-proxy, or may use a different artifact naming convention.

Historical Trend

This is the first token usage report — no prior data available for comparison. Future reports will show weekly cost trends and flag regressions.

References:

Generated by Daily Copilot Token Usage Analyzer ·

Metadata

Metadata

Assignees

No one assigned

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions