Building A Safe AI Assistant For Pull Request Summaries

This PR updates the billing logic and improves tests.

This PR changes payment retry behavior so recoverable gateway errors are retried up to 3 times, while hard declines are not retried.

Reviewer focus:
- Confirm the recoverable error list matches gateway documentation.
- Check retry count behavior in PaymentRetryService.
- Review the new tests for hard declines and max attempts.

Tests:
- Added PaymentRetryServiceTest.
- CI reports php artisan test passed.

Risk:
- Billing behavior changed.
- No database migration.

git diff --name-status origin/main...HEAD
git diff --stat origin/main...HEAD
git diff --unified=80 origin/main...HEAD

type ChangedFile = {
  path: string;
  status: 'added' | 'modified' | 'deleted' | 'renamed';
  additions: number;
  deletions: number;
};

function classifyFile(path: string): string[] {
  const tags: string[] = [];

  if (path.includes('migrations/')) tags.push('database_migration');
  if (path.includes('routes/')) tags.push('route_change');
  if (path.includes('auth') || path.includes('permissions')) tags.push('auth');
  if (path.includes('billing') || path.includes('payment')) tags.push('billing');
  if (path.includes('config/')) tags.push('configuration');
  if (path.includes('tests/')) tags.push('test');
  if (path.endsWith('.yml') || path.endsWith('.yaml')) tags.push('ci_or_config');

  return tags;
}

- if ($attempts > 3) {
+ if ($attempts >= 3) {
    return false;
  }

Updated retry condition in PaymentRetryService.

The service now stops retrying at attempt 3 instead of allowing the third retry attempt.

You are summarizing a pull request for human reviewers.

Do not only list changed files.
Explain the behavior that changed.
Base your summary only on the diff and provided CI/test data.
If behavior is unclear, say it is unclear.
Do not claim tests were run unless test results are provided.

type Risk = {
  level: 'low' | 'medium' | 'high';
  reason: string;
  file: string;
};

function detectRisk(file: ChangedFile): Risk[] {
  const tags = classifyFile(file.path);
  const risks: Risk[] = [];

  if (tags.includes('database_migration')) {
    risks.push({
      level: 'high',
      file: file.path,
      reason: 'Database migration can affect production data and deployment order.',
    });
  }

  if (tags.includes('auth')) {
    risks.push({
      level: 'high',
      file: file.path,
      reason: 'Authentication or authorization behavior changed.',
    });
  }

  if (tags.includes('billing')) {
    risks.push({
      level: 'high',
      file: file.path,
      reason: 'Billing or payment behavior changed.',
    });
  }

  if (tags.includes('configuration')) {
    risks.push({
      level: 'medium',
      file: file.path,
      reason: 'Configuration change may affect runtime behavior.',
    });
  }

  return risks;
}

Tests found in diff:
- Added tests/Feature/Billing/PaymentRetryServiceTest.php

Tests reported by CI:
- php artisan test: passed
- vendor/bin/phpstan analyse: passed

Potential missing tests:
- No test found for retry behavior when attempts = 3.
- No test found for unknown gateway error codes.

{
  "ci": {
    "status": "passed",
    "checks": [
      {
        "name": "php artisan test",
        "status": "passed",
        "duration_seconds": 84
      },
      {
        "name": "phpstan",
        "status": "passed",
        "duration_seconds": 31
      }
    ]
  }
}

Only list a test under "Tests run" if it appears in the CI data or user-provided command output.
If test data is missing, write: "No test run information was provided."

Schema::table('orders', function (Blueprint $table) {
    $table->index(['user_id', 'created_at']);
});

Migration note:
- Adds a compound index on orders(user_id, created_at).
- This can improve queries that filter by user and sort/filter by creation date.
- On large orders tables, index creation may be slow or require an online migration strategy depending on the database engine and deployment setup.

import { z } from 'zod';

const PullRequestSummarySchema = z.object({
  overview: z.string(),
  changedBehavior: z.array(z.string()),
  reviewerFocus: z.array(z.string()),
  riskyFiles: z.array(z.object({
    file: z.string(),
    reason: z.string(),
    level: z.enum(['low', 'medium', 'high']),
  })),
  testsFound: z.array(z.string()),
  testsRun: z.array(z.string()),
  missingTests: z.array(z.string()),
  migrationNotes: z.array(z.string()),
  unknowns: z.array(z.string()),
});

function renderSummary(summary: z.infer<typeof PullRequestSummarySchema>): string {
  return `
## Summary

${summary.overview}

## Changed behavior
${summary.changedBehavior.map(item => `- ${item}`).join('\n')}

## Reviewer focus
${summary.reviewerFocus.map(item => `- ${item}`).join('\n')}

## Risky files
${summary.riskyFiles.map(risk => `- **${risk.level}**: \`${risk.file}\` — ${risk.reason}`).join('\n')}

## Tests

**Tests found:**
${summary.testsFound.map(item => `- ${item}`).join('\n') || '- None found'}

**Tests run:**
${summary.testsRun.map(item => `- ${item}`).join('\n') || '- No test run information was provided'}

**Potential missing tests:**
${summary.missingTests.map(item => `- ${item}`).join('\n') || '- None detected'}

## Migration notes
${summary.migrationNotes.map(item => `- ${item}`).join('\n') || '- No migration notes'}

## Unknowns
${summary.unknowns.map(item => `- ${item}`).join('\n') || '- None'}
`;
}

Unknowns:
- The diff changes retry behavior, but no gateway documentation was provided to confirm the recoverable error list.
- No CI result was provided, so tests run cannot be verified.
- The migration adds an index to orders, but table size is unknown.

This PR is safe to merge.

No blocker was detected from the provided diff and CI data, but human review is still required.

All edge cases are covered.

The diff includes tests for the main success path and max retry count. No test was found for unknown gateway error codes.

# AI PR Summary Instructions

For this repository:

- Treat changes under app/Billing as high risk.
- Treat changes under app/Auth as high risk.
- Mention database migrations clearly.
- Mention queue/job changes because they may affect async behavior.
- Do not say tests passed unless CI data confirms it.
- Always include reviewer focus.
- Keep the summary concise and technical.

Generated summary based on the PR diff and available CI data. Please verify behavior and risk before merging.

1. Pull request opens.
2. System collects changed files, diff stats, and CI results.
3. Deterministic classifier marks risky areas.
4. AI receives curated diff context, risk metadata, and repository instructions.
5. AI returns structured JSON.
6. System validates JSON schema.
7. System renders Markdown summary.
8. Summary is posted as a PR comment or used to update the PR description.
9. Human reviewer reviews the PR.