FixSense — AI Test Failure Analysis

Stop debugging CI failures manually. FixSense analyzes your failing tests with AI, explains the root cause, classifies app bugs vs test bugs, scores flakiness, and suggests fixes — all as a PR comment.

Works with any test framework that produces JUnit XML reports.

What You Get

For every failing test, FixSense provides:

Feature	Description
Root Cause	AI explains why your test failed in 2-4 sentences
App Bug vs Test Bug	Is your application broken or just the test?
Flakiness Score	0-100 score — high means likely flaky, not a real regression
Suggested Fix	Actionable steps to resolve the failure
PR Comment	All results posted as a single, auto-updating PR comment

Quick Start

Add to your workflow after your test step:

- name: Run tests
  id: tests
  run: npx playwright test
  continue-on-error: true

- name: FixSense Analysis
  if: steps.tests.outcome == 'failure'
  uses: CopilotMe/fixsense-action@v1
  with:
    api-key: ${{ secrets.FIXSENSE_API_KEY }}

That's it. Next time a test fails, you'll get an analysis comment on your PR.

Full Workflow Example

name: Tests
on: [push, pull_request]

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: 20

      - run: npm ci
      - run: npx playwright install --with-deps

      - name: Run Playwright tests
        id: tests
        run: npx playwright test
        continue-on-error: true

      - name: FixSense Analysis
        if: steps.tests.outcome == 'failure'
        uses: CopilotMe/fixsense-action@v1
        with:
          api-key: ${{ secrets.FIXSENSE_API_KEY }}
          results-path: test-results/

      - uses: actions/upload-artifact@v4
        if: always()
        with:
          name: test-results
          path: test-results/

PR Comment Preview

The action posts a summary table with expandable details for each failure:

Test	Root Cause	Flakiness	Confidence
login.spec.ts > should login	Timeout waiting for selector .submit-btn — element no longer exists after recent refactor	🟡 65	🟢 high
checkout.spec.ts > payment	HTTP 500 from /api/payment — server error, not a test issue	🔴 10	🟢 high

Expanded details include:

• Full root cause explanation
• Suggested fix with code pointers
• Related code change (diff hint)
• Whether it's an app bug or test bug

Inputs

Input	Required	Default	Description
api-key	Yes	—	Your FixSense API key (get one free)
results-path	No	test-results/	Path to JUnit XML test results
post-comment	No	true	Post analysis as PR comment
api-url	No	https://fix-sense.vercel.app	API URL (for self-hosted)

Outputs

Output	Description
analyses	JSON array of analysis results
failure-count	Number of failures analyzed

Supported Test Frameworks

Any framework that produces JUnit XML reports:

Framework	Reporter Config
Playwright	reporter: [['junit', { outputFile: 'test-results/junit.xml' }]]
Cypress	cypress-junit-reporter
Jest	jest-junit
pytest	--junitxml=test-results/junit.xml
JUnit / TestNG	Native XML output
Mocha	mocha-junit-reporter
RSpec	rspec_junit_formatter
Go	go-junit-report
WebdriverIO	@wdio/junit-reporter — reporters: [['junit', { outputDir: './test-results' }]]
Dart / Flutter	dart test --machine | tojunit (junitreport)

Get Your API Key

1. Go to fix-sense.vercel.app
2. Sign in with GitHub or Google
3. Go to Settings and generate your API key
4. Add it as a repository secret named FIXSENSE_API_KEY

Free plan includes 350 analyses/month — enough for most small-to-mid teams.

Why FixSense?

• Saves 10-30 min per failure — no more scrolling through logs
• Catches flaky tests — high flakiness score = don't block the PR
• Distinguishes real bugs — HTTP 5xx, missing elements, broken flows flagged as app bugs
• Works with any CI — GitHub Actions, self-hosted runners, any JUnit XML producer
• Dashboard included — track failure trends, top flaky tests, auto-fix success rates at fix-sense.vercel.app

Links

• FixSense Dashboard
• Documentation
• GitHub App (zero-config alternative)

License

MIT