AI-Powered File Analysis
How CodeDD analyzes individual source files
AI-Powered File Analysis
CodeDD's file analysis stage reviews each scoped file for quality, security, and maintainability — surfacing issues that pattern-only scanners often miss.
Beyond traditional SAST
Traditional static analysis relies on pattern matching and syntax rules. CodeDD adds semantic understanding: business logic review, intent-based risk detection, and context from file type and role in the system.
Security findings go through a validation step that checks for supporting evidence before they affect your score. Optional SonarQube integration adds language-specific static rules when enabled.
What gets analyzed
Application code — business logic, APIs, database access, auth, validation, error handling.
Infrastructure — Dockerfiles, Kubernetes manifests, CI/CD configs, IaC templates.
Configuration — app settings, environment files, third-party integrations.
File selection
Deep analysis runs on files marked in audit scope. Priority goes to security-critical paths, core business logic, recently modified files, and high-complexity code.
Excluded: auto-generated code, minified files, binaries, and test fixtures.
Per-file assessment
Each analyzed file receives structured scores across dimensions you see in the dashboard:
| Area | What it covers |
|---|---|
| Code quality | Readability, consistency, modularity, maintainability, technical debt |
| Functionality | Completeness, edge cases, error handling |
| Performance | Efficiency, scalability, resource use |
| Security | Input validation, data handling, authentication |
| Standards | Best practices, design patterns, complexity |
Findings include severity (Green / Yellow / Orange / Red), confidence level, and specific remediation guidance.
Security findings
Common categories: injection flaws, XSS vectors, auth bypasses, insecure cryptography, exposed secrets, sensitive data handling.
Each security flag carries a confidence score from the validation pipeline. Findings at or above the 80% confidence threshold are treated as validated and actionable; lower-confidence items are marked inconclusive for manual review in the Security / Flags tab.
Dependency analysis
CodeDD scans package manifests and imports across major ecosystems (npm, pip, Maven, Go modules, Cargo, Composer, .NET, and others).
For each dependency:
- Version and known CVEs (NVD, GitHub Advisory, OSV)
- Severity and patch availability
- License type and compliance flags
Portfolio views include a Supply Chain Vulnerabilities panel and License Compliance section.
Complexity metrics
Cyclomatic complexity is calculated per function using Radon/Lizard. High-complexity functions (typically 20+) are flagged as refactoring candidates and contribute to technical debt signals.
SonarQube (optional)
When enabled, SonarScanner runs in an isolated container against a temporary workspace. Results are merged with AI findings. The workspace is deleted after analysis.
Data privacy
After analysis:
- Stored: file paths, metrics, findings, dependency lists, complexity scores
- Never stored: source code content, detected secrets (flagged but not persisted), PII from comments