Source code ↗

Flaky Test Analyzer

Analyze flaky test behavior from GitHub Actions artifacts and prioritize tests by wasted CI time and instability signals.

Developed by Vineet Kumar
Detected Flaky Tests
0
Total Wasted CI Time (s)
0.0
Highest Flake Score
0.00
Demo disclaimer
This web UI currently supports public GitHub repository URLs for quick evaluation and demo flows.
For private repositories, use the workflow at ci-snippet/flaky-tests.yml (or copy equivalent steps) in your repo CI pipeline, then run this analyzer against generated reports in your controlled environment.
Why this matters: reliable flake detection needs test reruns/repeats; single run per commit usually cannot expose flaky behavior.
Before you fetch: the target repository must upload JUnit/Surefire XML test artifacts in GitHub Actions (for example via an upload-artifact step). If no matching test artifact exists, the dashboard cannot analyze flakiness.
Tip: set GITHUB_TOKEN on the server for better API reliability and fewer rate-limit failures.
AI Explain setup (model-agnostic)
Flaky analysis works without any LLM. Explain supports LLM_PROVIDER=anthropic|openai|ollama.
Examples: LLM_PROVIDER=anthropic + ANTHROPIC_API_KEY, or LLM_PROVIDER=openai + OPENAI_API_KEY, or LLM_PROVIDER=ollama.
Optional model override: LLM_MODEL (for example gpt-4o-mini or llama3.2).
No repo handy? Seed realistic flaky tests to try the dashboard + AI explain.

No flaky tests yet. Fetch a repo or load demo data above (real detection needs more than one run per commit, e.g. reruns).