# Changelog of GitHub Repo Search — Stars, Language & Topics (`ryanclinton/github-repo-search`) Actor

- **URL**: https://apify.com/ryanclinton/github-repo-search/changelog.md
- **Full Actor documentation**: https://apify.com/ryanclinton/github-repo-search.md

## Changelog

### 2.3 (2026-09-04) — Fix: results now repeat. Two identical runs used to disagree

- Two runs of the same search minutes apart could return different scores, different verdicts and a different order. One repository moved 36 places between runs; nine adoption verdicts changed. The cause was repository statistics that GitHub had not finished preparing being given up on too early, so each run saw a different subset. The actor now waits properly for them
- Run-to-run change detection is affected by the same fix: comparisons against a previous run were reporting statistics gaps as if they were real changes in the repository
- The run summary now includes a scoringCoverage section stating how many scored rows were missing each data surface, so a score computed partly from estimates is visible rather than implied
- Repositories with an unknown contributor count are no longer scored as though they had the fewest possible contributors

### 2.3 (2026-09-04) — Fix: repositories with many contributors were reported as single-maintainer

- When GitHub had not finished computing a repository's contributor statistics, the actor reported that repository as having exactly one contributor holding 100% of commits, and flagged it as a single-maintainer bus-factor risk. This affected roughly 13% of enriched rows, including large well-staffed projects. Contributor count is now reported as unavailable in that case, and scoring treats it as unknown rather than as a one-person project
- The adoption verdict for a repository is now the same whichever solution mode you run

### 2.3 (2026-09-04) — Accurate run reporting when comparing specific repositories

- Comparing a list of repositories no longer reports a search query you never entered. The run status, the log, and the run summary now say which repositories were compared, and the summary records them under comparedRepos
- The upfront cost estimate for a comparison run is now based on the number of repositories you listed, instead of the unrelated maximum-results setting
- Score explanations no longer describe an abandoned repository as active. The archive check now says only whether the repository is archived on GitHub
- Where commit-activity data is unavailable for a repository, the score explanation now states that plainly rather than suggesting a setting change that would not help
- Documentation: enrichment is on by default, and README text is searchable with in:readme

### 2.3 (2026-08-27) — Clearer warnings and accurate time estimates on large runs without a GitHub token

- Large runs without a githubToken now say so in the run status before any work starts, with a link to create one, instead of only mentioning it in the log
- The time remaining shown during enrichment is now measured from the run's actual pace instead of an assumed one, so a slow run reports minutes rather than seconds
- A run being slowed by GitHub's anonymous limits now says that in the run status, with the fix, rather than appearing stuck

### 2.3 (2026-06-19) — Faster pay-per-event delivery, accurate mode presets, upfront cost estimate

- Pay-per-event runs now bill once at the end instead of once per repository — the final step finishes faster on large runs, with no change to what you are charged.
- Solution modes now correctly apply their own sort and filter presets. Previously the manual default settings silently overrode the mode you selected.
- The run status now shows a worst-case cost estimate up front for pay-per-event runs, so the maximum is visible before any charge.

### 2.3 (2026-06-09) — Deliver results to Slack and Notion

- Send each run's intelligence digest (top repositories, scores, and monitoring changes) straight into your own Slack or Notion using Apify MCP connectors. Pair with "Compare to previous run" on a schedule for a standing GitHub watch that posts only what changed. Your Slack/Notion credentials stay on your Apify account — the actor never sees them. Leave the new connector fields blank and the actor works exactly as before.

### 2.3 (2026-05-23) — Fix: rare hang when persistent state store is unavailable

### 2.3 (2026-05-16) — Reliability: partial enrichment when run budget is tight

### 2.2 (2026-05-15) — Intelligence on by default

### 2.1 (2026-05-10) — Header-aware rate-limit retry

### 2.0 (2026-04-07) — v2.0: GitHub Repo Intelligence — scoring, modes, and monitoring

### 1.2 (2026-04-07) — Performance optimization, new features, and UX improvements

### 1.1 (2026-03-29) — Fix schema validation and improve reliability