[deep-report] Investigate 22K GitHub REST API calls/day: optimize Smoke suite and Changeset Generator (68% of quota)

[github-actions]

Problem Statement

The new GitHub API Consumption Report (2026-05-06, #30598) reveals that the Smoke test suite + Changeset Generator consume ~14,975 of 22,029 daily REST API calls (68%), with individual workflows like Smoke Gemini (2,961 calls/run) and Changeset Generator (1,488 calls/run) as the top consumers.

With only one day of baseline data, there's no historical context yet — but the day-1 pattern clearly shows the Smoke suite as the primary optimization target.

Expected Impact

• A 20% reduction in Smoke suite API calls would save ~3,000 calls/day
• Reduces risk of hitting the 15,000/hr per-workflow rate limit during peak activity
• Enables more headroom for new agent workflows without quota pressure

Investigation Steps

1. Audit what GitHub API calls the Smoke test workflows are making per run (using agenticworkflows audit on a recent Smoke Gemini run)
2. Identify redundant or batched-able calls (e.g., fetching the same PR data multiple times)
3. Check if the Changeset Generator is re-fetching repository data it could cache
4. Explore using GraphQL batching instead of multiple REST calls where possible

Estimated Effort

Medium (2-4 hours investigation + patch)

Data Source

GitHub API Consumption Report — 2026-05-06 (#30598); first baseline data point

Generated by DeepReport - Intelligence Gathering Agent

Generated by DeepReport - Intelligence Gathering Agent · ● 330.7K · ◷

• expires on May 8, 2026, 3:35 PM UTC