Private/Test-AICommitPattern.ps1
|
# Core AI Detection Logic - Private Functions # Extracted from LLMCommitIdentifier.ps1 and enhanced for module use function Test-AICommitPattern { <# .SYNOPSIS Core AI detection algorithm with sophisticated pattern recognition. .DESCRIPTION Analyzes commit messages and metadata using advanced pattern matching to identify AI assistance likelihood. Implements 0-11+ scoring system with multiple detection categories. .PARAMETER Commits Array of commit objects to analyze .PARAMETER ShowDetails Display detailed scoring information #> [CmdletBinding()] param( [Parameter(Mandatory)] [array]$Commits, [switch]$ShowDetails ) # Enhanced AI pattern detection arrays $aiIndicators = @{ "High Probability" = @( "comprehensive", "implementation", "integration", "sophisticated", "systematic", "optimization", "refactoring", "enhancement", "architecture", "infrastructure", "methodology", "framework" ) "Medium Probability" = @( "setup", "configure", "enhance", "update", "create", "improve", "initialize", "establish", "develop", "construct", "build" ) "AI Tools Mentioned" = @( "copilot", "claude", "ai", "assistant", "llm", "chatgpt", "gpt-4", "cursor", "tabnine", "codewhisperer", "intellicode", "codeium" ) "Technical Patterns" = @( "typescript", "eslint", "prettier", "webpack", "babel", "jest", "dockerfile", "kubernetes", "cicd", "pipeline", "automation" ) "Documentation Patterns" = @( "readme", "documentation", "comments", "jsdoc", "swagger", "api-docs", "changelog", "contributing", "license", "usage-examples" ) } # Conventional commit pattern (strong AI indicator) $conventionalCommitPattern = "^(feat|fix|docs|style|refactor|test|chore|build|ci|perf|revert)(\(.+\))?: .+" $analysisResults = @() foreach ($commit in $Commits) { $aiScore = 0 $indicators = @() $commitText = "$($commit.message) $($commit.author)".ToLower() if ($ShowDetails) { Write-Host "Analyzing commit: $($commit.hash)" -ForegroundColor Gray } # Category-based pattern detection with scoring foreach ($category in $aiIndicators.Keys) { $categoryHits = 0 foreach ($pattern in $aiIndicators[$category]) { if ($commitText -match [regex]::Escape($pattern)) { $categoryHits++ # Scoring based on category importance switch ($category) { "High Probability" { $aiScore += 3 } "Medium Probability" { $aiScore += 2 } "AI Tools Mentioned" { $aiScore += 5 } "Technical Patterns" { $aiScore += 1.5 } "Documentation Patterns" { $aiScore += 1 } } $indicators += "$category`: $pattern" if ($ShowDetails) { Write-Host " Match: $category -> $pattern" -ForegroundColor Yellow } } } # Bonus for multiple hits in same category (suggests AI comprehensiveness) if ($categoryHits -gt 2) { $aiScore += 1 $indicators += "$category`: Multiple Patterns (+1 bonus)" } } # Conventional commit format detection if ($commit.message -match $conventionalCommitPattern) { $aiScore += 2 $indicators += "Conventional Commit Format" if ($ShowDetails) { Write-Host " Conventional commit detected (+2)" -ForegroundColor Green } } # Message length analysis (AI tends to be comprehensive) $messageLength = $commit.message.Length if ($messageLength -gt 50) { $lengthBonus = [Math]::Min(2, [Math]::Floor($messageLength / 50)) $aiScore += $lengthBonus $indicators += "Detailed Message Length: $messageLength chars (+$lengthBonus)" } # Timestamp pattern analysis (AI often used during intense coding sessions) if ($commit.date) { $commitTime = [DateTime]::Parse($commit.date) $hour = $commitTime.Hour # AI assistance often used during late night coding (higher likelihood) if (($hour -ge 22) -or ($hour -le 6)) { $aiScore += 0.5 $indicators += "Late Night Coding Pattern (+0.5)" } # Rapid commits suggest AI assistance # (This would need previous commit for comparison - simplified here) if ($commit.message -match "(rapid|quick|fast|immediate)") { $aiScore += 1 $indicators += "Rapid Development Indicator (+1)" } } # File type analysis (some file types more likely to use AI) if ($commit.message -match "\.(ts|js|py|cs|java|cpp|h|md|json|yml|yaml|dockerfile)") { $aiScore += 0.5 $indicators += "AI-Friendly File Types (+0.5)" } # Calculate AI likelihood based on score $aiLikelihood = switch ($aiScore) { { $_ -ge 8 } { "Very High" } { $_ -ge 5 } { "High" } { $_ -ge 3 } { "Medium" } { $_ -ge 1 } { "Low" } default { "Minimal" } } # Determine most likely AI model based on patterns $likelyModel = "Human" if ($aiScore -ge 3) { if ($commitText -match "copilot") { $likelyModel = "GitHub Copilot" } elseif ($commitText -match "claude") { $likelyModel = "Claude AI" } elseif ($commitText -match "(chatgpt|gpt-4|openai)") { $likelyModel = "ChatGPT" } elseif ($commitText -match "cursor") { $likelyModel = "Cursor AI" } else { $likelyModel = "AI Assistant (Generic)" } } if ($ShowDetails) { Write-Host " Final Score: $aiScore -> $aiLikelihood" -ForegroundColor Cyan } # Create analysis result $analysisResults += [PSCustomObject]@{ Hash = $commit.hash ShortHash = $commit.short Message = $commit.message Author = $commit.author Date = $commit.date AIScore = [Math]::Round($aiScore, 1) AILikelihood = $aiLikelihood Indicators = $indicators PossibleLLM = $likelyModel } } return $analysisResults } |