AI Contribution Reporter

AI Contribution Reporter is a production-oriented Visual Studio Code extension that tracks coding activity in real time and classifies AI versus human code contribution before any commit happens.

Key Capabilities

• Real-time tracking for file create/delete/rename/modify/save events
• Real-time activity inference for manual typing, multi-line insertion, replacement, copy/paste, refactoring, and AI completion acceptance signals
• Contribution classification into:
  • AI_GENERATED
  • HUMAN_WRITTEN
  • AI_MODIFIED_HUMAN_CODE
  • HUMAN_MODIFIED_AI_CODE
  • UNKNOWN
• File-level metrics and workspace-level ownership summaries
• Dashboard in Activity Bar with cards and Chart.js visualizations (pie, bar, trend)
• Alerts when AI contribution exceeds threshold or human contribution drops below threshold
• Multi-format report export: HTML, JSON, CSV, Markdown
• Local-first persistence in extension global storage (JSON database)
• No dependency on Git history, commit history, pull requests, or server-side processing

Activity Bar Views

• Dashboard
• File Analysis
• Developer Analysis
• Reports
• Settings

Commands

• AI Reporter: Generate Report
• AI Reporter: Open Dashboard
• AI Reporter: Export HTML Report
• AI Reporter: Export JSON Report
• AI Reporter: Export CSV Report
• AI Reporter: Refresh Metrics
• AI Reporter: Reset Workspace Metrics

Settings

• aiReporter.enableTracking
• aiReporter.enableAIDetection
• aiReporter.enableHumanDetection
• aiReporter.autoGenerateReports
• aiReporter.reportOutputFolder
• aiReporter.includeExtensions
• aiReporter.excludeFolders
• aiReporter.dataRetentionDays
• aiReporter.aiContributionThreshold
• aiReporter.humanContributionThreshold

Architecture

See docs/ARCHITECTURE.md for full architecture and service responsibilities.

Development

npm install
npm run compile
npm run test

Press F5 in VS Code to launch the extension host.

Notes on Classification Accuracy

The extension performs pre-commit, editor-native inference. AI detection accuracy depends on the availability of command execution signals and editing patterns. Confidence scores are included to support governance decisions.