GitHub Actions Analyzer

Analyze GitHub Actions performance and visualize workflow timelines directly in your terminal or Perfetto UI.

Install

go install github.com/stefanpenner/gha-analyzer/cmd/gha-analyzer@latest

Authentication

If you have the GitHub CLI installed and authenticated (gh auth login), the token is picked up automatically — no extra setup needed.

Otherwise, set GITHUB_TOKEN:

export GITHUB_TOKEN="your_token_here"

Usage

# Analyze a PR or Commit
gha-analyzer nodejs/node/pull/60369

# Save and open in Perfetto UI
gha-analyzer <url> --perfetto=trace.json --open-in-perfetto

# Filter to recent activity
gha-analyzer <url> --window=1h

Historical Trend Analysis

Analyze workflow performance trends over time to identify patterns, flaky jobs, and performance regressions:

# Analyze trends for the last 30 days (default)
gha-analyzer trends owner/repo

# Analyze trends for a specific time period
gha-analyzer trends owner/repo --days=7

# Filter by branch (e.g., only main branch workflows)
gha-analyzer trends owner/repo --branch=main

# Filter by workflow file (e.g., only post-merge workflows)
gha-analyzer trends owner/repo --workflow=post-merge.yaml

# Combine filters for focused analysis
gha-analyzer trends owner/repo --days=14 --branch=main --workflow=ci.yaml

# Export as JSON for programmatic analysis
gha-analyzer trends owner/repo --days=14 --format=json

Trend analysis provides:

• Summary Statistics: Average, median, and 95th percentile durations
• Success Rate Trends: Track workflow reliability over time
• Duration Trends: Visualize performance changes with ASCII charts
• Job Performance: Identify slowest jobs and their trends
• Flaky Job Detection: Automatically detect jobs with inconsistent outcomes (>10% failure rate)
• Trend Direction: See if performance is improving, stable, or degrading

Sampling

For large repositories, fetching job details for every workflow run is expensive (1 API call per run). Trend analysis uses job-level statistical sampling to reduce API usage while maintaining accuracy.

How it works: All workflow runs are always fetched (cheap — 1 call per 100 runs), so run-level metrics (duration trends, success rates) are exact. Job detail fetching uses stratified temporal sampling — runs are bucketed into equal-width time intervals and samples are drawn proportionally from each bucket, guaranteeing even temporal coverage. Sample size is computed using a finite-population formula at 95% confidence with ±10% margin of error. For example, a repo with 1,000 runs in 30 days will fetch job details for ~88 runs instead of all 1,000.

What's affected: Job-level analysis (per-job trends, regressions, improvements, flaky detection, queue times) uses the sampled subset. Rare jobs with only a few runs may not appear in sampled results. Trend directions for borderline jobs may vary between runs.

What's not affected: Total run count, average/median/p95 duration, success rate, and overall trend direction are always computed from the full dataset.

# Default: sampling enabled (95% confidence, ±10% margin)
gha-analyzer trends owner/repo --days=30

# Disable sampling for exact results (more API calls)
gha-analyzer trends owner/repo --days=30 --no-sample

# Tune confidence and margin
gha-analyzer trends owner/repo --confidence=0.99 --margin=0.05

Sample Output

================================================================================
📈 Historical Trend Analysis: stefanpenner/gha-analyzer
================================================================================

Summary Statistics
------------------
Average Duration                        1m 46s
Median Duration                         1m 41s
95th Percentile                         3m 13s
Average Success Rate                     61.7%
Trend Direction           ✓ Improving (-20.7%)
Flaky Jobs Detected                          1

OTel Output

# JSON spans to stdout (agent-readable JSONL)
gha-analyzer <url> --otel

# Export via OTLP/HTTP
gha-analyzer <url> --otel=localhost:4318

# Export via OTLP/gRPC
gha-analyzer <url> --otel-grpc
gha-analyzer <url> --otel-grpc=localhost:4317

Webhook Input

Pipe a GitHub Actions webhook payload on stdin to analyze the associated commit:

echo '{"workflow_run":{"head_sha":"abc123"},"repository":{"full_name":"owner/repo"}}' | gha-analyzer --otel

Supports workflow_run and workflow_job events. When no URL arguments are given and stdin is piped, the webhook is read automatically.

Development

The project uses Bazel for hermetic builds.

# Run the analyzer
bazel run //:gha-analyzer -- <url>

# Build everything
bazel build //...

# Run all tests
bazel test //...

# Regenerate BUILD.bazel files after adding Go packages
bazel run //:gazelle

License

MIT