fractalyze/ZK Benchmark

Run benchmarks, detect regressions, and report results

View on GitHub

Trust Signals

Scorecard Score
not yet scored
Maintenance Recency
Activelast commit Jun 29, 2026
License
Apache 2.0

Pinned Snippet

workflow.ymlSHA-pinned
uses: fractalyze/benchmark-action@01b8a1e990471732d0dab25af13a14d52e697a85 # v10

tags can be moved; commit SHAs can't. why a SHA?

namedescriptionrequireddefault
benchmark_cmdCommand to run the benchmark (must output JSON to benchmark_results.json)yes
deviceDevice type (cpu or gpu)yes
regression_thresholdPercentage threshold fallback when stdev is unavailable (e.g., 0.10 for 10%)no0.10
zscore_thresholdZ-score threshold for regression detection (e.g., 3.0)no3.0
baseline_pathPath to baseline JSON file for regression detectionnobenchmark_data/baseline.json
results_filePath to output benchmark results JSONnobenchmark_results.json
slack_webhookSlack webhook URL for regression alerts (fallback if no bot token)no""
slack_bot_tokenSlack Bot Token (xoxb-...) for thread replies. Preferred over slack_webhook.no""
slack_channel_idSlack channel ID for bot messages (e.g., C0123456789)no""
store_resultsWhether to store results (true on main branch)nofalse
results_dirDirectory to store historical resultsnobenchmark_data
cpu_load_thresholdCPU load threshold (0-1) above which to emit warningno0.80
memory_thresholdMemory usage threshold (0-1) above which to emit warningno0.80
rolling_windowNumber of historical results to average for baselineno10
anthropic_api_keyAnthropic API key for Claude AI analysisno""
ai_modelClaude model to use for analysisnoclaude-opus-4-5-20251101
dashboard_tokenGitHub token for triggering dashboard updatesno""
dashboard_repoDashboard repository (owner/repo)nofractalyze/benchmark-dashboard
repoSource repository name for commit links (e.g., zkx). Defaults to GITHUB_REPOSITORY tail.no""
namedescription
has_significant_changeWhether a significant change (regression or improvement) was detected
change_typeType of change detected (regression, improvement, mixed, or empty)
results_filePath to benchmark results JSON
cpu_loadNormalized CPU load (0-1)
memory_usageMemory usage ratio (0-1)