Skip to main content

OpenAI Integration

x-fidelity offers advanced AI-powered code analysis through integration with OpenAI's language models. This feature provides in-depth insights and suggestions for improving your codebase.

Overview​

The OpenAI integration:

  • Analyzes code patterns and practices
  • Identifies potential issues
  • Suggests improvements
  • Provides severity ratings
  • Supports custom prompts

Configuration​

Environment Variables​

# Required
OPENAI_API_KEY=your_api_key_here

# Optional
OPENAI_MODEL=gpt-4  # Default is 'gpt-4'

CLI Options​

Enable OpenAI analysis:

xfidelity . -o true

Creating OpenAI Rules​

Rules using OpenAI must:

  • Have names starting with 'openai'
  • Use the openaiAnalysis fact
  • Use the openaiAnalysisHighSeverity operator

Example rule:

{
    "name": "openaiAnalysisA11y-global",
    "conditions": {
        "all": [
            {
                "fact": "fileData",
                "path": "$.fileName",
                "operator": "equal",
                "value": "REPO_GLOBAL_CHECK"
            },
            {
                "fact": "openaiAnalysis",
                "params": {
                    "prompt": "Identify accessibility issues",
                    "resultFact": "openaiAnalysisA11y"
                },
                "operator": "openaiAnalysisHighSeverity",
                "value": 7
            }
        ]
    },
    "event": {
        "type": "warning",
        "params": {
            "message": "OpenAI detected accessibility issues",
            "details": {
                "fact": "openaiAnalysisA11y"
            }
        }
    }
}

Response Format​

OpenAI responses follow this structure:

{
    "issues": [
        {
            "issue": "Missing alt text",
            "severity": 8,
            "description": "Images lack alt text for screen readers",
            "filePaths": ["src/components/Image.tsx"],
            "suggestion": "Add alt attributes to all img elements",
            "codeSnippets": [
                {
                    "filePath": "src/components/Image.tsx",
                    "lineNumber": 15,
                    "before": "<img src=\"logo.png\" />",
                    "after": "<img src=\"logo.png\" alt=\"Company logo\" />"
                }
            ]
        }
    ]
}

Best Practices​

  1. Cost Management:

    • Use specific prompts
    • Limit analysis scope
    • Monitor API usage
    • Cache results when possible

  2. Security:

    • Never send sensitive data
    • Review code before sending
    • Use data masking
    • Set rate limits

  3. Performance:

    • Use async analysis
    • Implement timeouts
    • Handle API errors
    • Cache responses

  4. Prompts:

    • Be specific
    • Focus on one aspect
    • Include context
    • Set clear expectations

Example Rules​

Security Analysis​

{
    "name": "openaiSecurityCheck-global",
    "conditions": {
        "all": [
            {
                "fact": "openaiAnalysis",
                "params": {
                    "prompt": "Identify security vulnerabilities",
                    "resultFact": "securityAnalysis"
                },
                "operator": "openaiAnalysisHighSeverity",
                "value": 9
            }
        ]
    },
    "event": {
        "type": "fatality",
        "params": {
            "message": "Critical security issues detected",
            "details": {
                "fact": "securityAnalysis"
            }
        }
    }
}

Performance Analysis​

{
    "name": "openaiPerformance-global",
    "conditions": {
        "all": [
            {
                "fact": "openaiAnalysis",
                "params": {
                    "prompt": "Identify performance bottlenecks",
                    "resultFact": "performanceAnalysis"
                },
                "operator": "openaiAnalysisHighSeverity",
                "value": 7
            }
        ]
    },
    "event": {
        "type": "warning",
        "params": {
            "message": "Performance issues detected",
            "details": {
                "fact": "performanceAnalysis"
            }
        }
    }
}

Limitations​

  1. Cost: API usage incurs charges
  2. Privacy: Code is sent to OpenAI
  3. Rate Limits: API has request limits
  4. Response Time: Analysis adds processing time
  5. Accuracy: Results may need human review

Next Steps​