Automatically analyze failing tests and suggest or apply fixes. (v6.2.0)

The Problem

Tests break constantly:

Selectors change when UI updates
Timing issues cause flaky failures
Page content changes break assertions
Elements move or get renamed

Manually fixing tests is tedious and time-consuming.

The Solution

CBrowser's AI Test Repair:

Runs your tests
Analyzes failures
Suggests specific fixes
Optionally applies them automatically
Verifies the repairs work

Quick Start

# Analyze a broken test and see suggestions
npx cbrowser repair-tests broken-test.txt

# Automatically apply repairs
npx cbrowser repair-tests tests.txt --auto-apply

# Apply repairs and verify they work
npx cbrowser repair-tests tests.txt --auto-apply --verify

# Save repaired tests to new file
npx cbrowser repair-tests tests.txt --auto-apply --output fixed-tests.txt

What It Detects

Failure Type	Detection	Repair Strategy
`selector_not_found`	Element doesn't exist	Find alternative selectors on current page
`assertion_failed`	Verify statement false	Suggest updated assertions based on actual content
`timeout`	Step took too long	Add explicit wait statements
`element_not_interactable`	Element hidden/disabled	Add scroll/wait before interaction

Example Output

🔧 AI TEST REPAIR REPORT
══════════════════════════════════════════════════════════════

📋 Analyzing: Login Tests

──────────────────────────────────────────────────────────────
TEST: Login Flow
──────────────────────────────────────────────────────────────

  → go to https://example.com
    ✓ Passed

  → click the signin button
    ✗ Failed: Failed to click: signin button

    💡 SUGGESTIONS:

    1. Update selector to "Login" (confidence: 70%)
       Original: click the signin button
       Fixed:    click "Login"
       Reason:   Found button with text "Login" on page

    2. Update selector to "Sign In" (confidence: 60%)
       Original: click the signin button
       Fixed:    click "Sign In"
       Reason:   Found link with text "Sign In" on page

    3. Add wait before click (confidence: 50%)
       Original: click the signin button
       Fixed:    wait 2 seconds
                 click the signin button
       Reason:   Element may not be loaded yet

  → type "[email protected]" in email
    ✓ Passed (skipped - previous step failed)

──────────────────────────────────────────────────────────────
TEST: Search
──────────────────────────────────────────────────────────────

  → verify page contains "10 results"
    ✗ Failed: Expected "10 results", found "12 results"

    💡 SUGGESTIONS:

    1. Update assertion (confidence: 90%)
       Original: verify page contains "10 results"
       Fixed:    verify page contains "12 results"
       Reason:   Page now shows "12 results"

    2. Use flexible assertion (confidence: 85%)
       Original: verify page contains "10 results"
       Fixed:    verify page contains "results"
       Reason:   Makes test resilient to count changes

══════════════════════════════════════════════════════════════
📊 SUMMARY
══════════════════════════════════════════════════════════════

  Tests analyzed: 2
  Failed steps: 2
  Suggestions generated: 5
  Auto-applicable: 3

  Run with --auto-apply to fix automatically

Auto-Apply Mode

When you use --auto-apply, CBrowser:

Selects the highest-confidence suggestion for each failure
Applies the fix to your test
Shows what changed

npx cbrowser repair-tests tests.txt --auto-apply

🔧 AUTO-APPLYING REPAIRS
══════════════════════════════════════════════════════════════

TEST: Login Flow

  Step 2: click the signin button
  ─────────────────────────────────
  - click the signin button
  + click "Login"

  ✓ Applied (confidence: 70%)

TEST: Search

  Step 1: verify page contains "10 results"
  ─────────────────────────────────
  - verify page contains "10 results"
  + verify page contains "results"

  ✓ Applied (confidence: 85%)

══════════════════════════════════════════════════════════════
📊 REPAIRS APPLIED: 2

Repaired test saved to: tests.txt.repaired

Verify Repairs

Add --verify to run the repaired tests and confirm they pass:

npx cbrowser repair-tests tests.txt --auto-apply --verify

🔧 VERIFYING REPAIRS
══════════════════════════════════════════════════════════════

Running repaired tests...

TEST: Login Flow
  ✓ go to https://example.com
  ✓ click "Login"              ← REPAIRED
  ✓ type "[email protected]" in email
  ✓ click submit
  ✅ PASSED

TEST: Search
  ✓ go to https://example.com/search
  ✓ type "query" in search box
  ✓ verify page contains "results"  ← REPAIRED
  ✅ PASSED

══════════════════════════════════════════════════════════════
📊 VERIFICATION: 2/2 tests now passing

API Usage

import {
  parseNLTestSuite,
  repairTestSuite,
  formatRepairReport,
  exportRepairedTest
} from 'cbrowser';

// Parse test content
const suite = parseNLTestSuite(testContent, "My Tests");

// Run repair analysis
const result = await repairTestSuite(suite, {
  autoApply: true,
  verifyRepairs: true,
  headless: true,
});

// Show report
console.log(formatRepairReport(result));

// Access repaired tests
for (const testResult of result.testResults) {
  if (testResult.wasRepaired) {
    console.log(`Repaired: ${testResult.testName}`);
    console.log(exportRepairedTest(testResult));
  }
}

// Get repair statistics
console.log(`Repair success rate: ${result.summary.repairSuccessRate}%`);

How the AI Works

Failure Analysis: Parses error messages to understand what went wrong
Page Inspection: Examines current page to find alternatives
Pattern Matching: Compares failed selector against available elements
Confidence Scoring: Ranks suggestions by likelihood of success
Validation: Optionally runs repaired test to confirm fix

The AI doesn't just guess—it examines the actual page state and makes informed suggestions.

Configuration

{
  "repair": {
    "autoApplyThreshold": 70,
    "maxSuggestionsPerStep": 3,
    "preferFlexibleAssertions": true,
    "addWaitsForTimeouts": true
  }
}

Option	Default	Description
`autoApplyThreshold`	70	Minimum confidence to auto-apply
`maxSuggestionsPerStep`	3	How many suggestions to show
`preferFlexibleAssertions`	true	Prefer resilient assertions
`addWaitsForTimeouts`	true	Add waits for timeout failures

Best Practices

Review before applying - Run without --auto-apply first
Use --verify - Confirm repairs actually work
Commit repaired tests - Keep your test suite up to date
Check confidence scores - Low confidence = review manually
Combine with flaky detection - Fix flaky tests first

Natural Language Tests - The test format this repairs
Flaky Test Detection - Find tests that need repair
CLI Reference - All repair-tests options

Copyright: (c) 2026 Alexa Eden.

License: MIT License

Contact: [email protected]

AI Test Repair