COR-Matrix Integration

COR-Matrix (Code Origin Retention Matrix) is an advanced analytics feature that tracks AI-generated code retention patterns in your codebase. It helps you understand how much AI-generated code persists over time and analyze the quality and utility of AI contributions.

Overview

COR-Matrix provides:

Code Origin Tracking: Identify which code was AI-generated vs human-written
Retention Analysis: Track how long AI-generated code remains in the codebase
Quality Metrics: Analyze patterns in code modifications and refactoring
Team Insights: Understand AI adoption and effectiveness across your team

How It Works

Code Origin Tagging

When HAI Build generates or modifies code:

Each change is tagged with metadata (timestamp, task ID, file path)
Tags are sent to the COR-Matrix service
The service tracks the code’s lifecycle
Analytics are generated on retention patterns

Retention Tracking

COR-Matrix monitors:

Initial Generation: When AI creates new code
Modifications: When humans edit AI-generated code
Deletions: When AI-generated code is removed
Persistence: How long code remains unchanged

Analytics Dashboard

Visualize:

Retention rates over time
Most stable AI-generated components
Frequently modified sections
Team adoption metrics

Configuration

Setup via .hai.config

Create or edit .hai.config in your workspace root:

touch .hai.config

Add COR-Matrix configuration:

# COR-Matrix Configuration
cormatrix.baseURL=https://cormatrix.example.com
cormatrix.token=your-api-token-here
cormatrix.workspaceId=your-workspace-id

Configuration Parameters

Parameter	Description	Required
`cormatrix.baseURL`	COR-Matrix service endpoint	Yes
`cormatrix.token`	Authentication token	Yes
`cormatrix.workspaceId`	Unique workspace identifier	Yes

Example Configuration

# .hai.config

# HAI Build Configuration
name=MyProject

# Telemetry (optional)
langfuse.apiUrl=https://langfuse.example.com
langfuse.apiKey=lf_key_xxx
langfuse.publicKey=lf_pub_xxx

posthog.url=https://posthog.example.com
posthog.apiKey=phc_xxx

# COR-Matrix Integration
cormatrix.baseURL=https://cormatrix.example.com/api
cormatrix.token=cm_token_xxx
cormatrix.workspaceId=workspace-123

Security Considerations

Protecting Sensitive Data

The .hai.config file is not git-ignored by default. Ensure you don’t commit sensitive tokens to version control.

Add to .gitignore:

echo ".hai.config" >> .gitignore

Environment-Specific Configuration

Use different configs for different environments: Local Development (.hai.config):

cormatrix.baseURL=http://localhost:8080
cormatrix.token=dev-token
cormatrix.workspaceId=dev-workspace

CI/CD Pipeline: Inject via environment variables or CI secrets:

# .github/workflows/ci.yml
- name: Configure COR-Matrix
  run: |
    cat > .hai.config << EOF
    cormatrix.baseURL=${{ secrets.CORMATRIX_URL }}
    cormatrix.token=${{ secrets.CORMATRIX_TOKEN }}
    cormatrix.workspaceId=${{ secrets.CORMATRIX_WORKSPACE }}
    EOF

Token Management

Rotate tokens regularly:

Generate new token in COR-Matrix dashboard
Update .hai.config locally
Update secrets in CI/CD platform
Revoke old token

Data Collected

Code Change Events

For each AI-generated change:

{
  "eventType": "code_generated",
  "timestamp": "2024-01-15T10:30:00Z",
  "taskId": "task-abc123",
  "file": "src/components/Button.tsx",
  "linesAdded": 45,
  "linesModified": 0,
  "linesDeleted": 0,
  "aiModel": "claude-sonnet-4-5-20250929",
  "userId": "user-xyz",
  "workspaceId": "workspace-123"
}

Modification Events

When AI-generated code is modified:

{
  "eventType": "code_modified",
  "timestamp": "2024-01-16T14:20:00Z",
  "originalTaskId": "task-abc123",
  "file": "src/components/Button.tsx",
  "linesModified": 5,
  "modifiedBy": "human",
  "retentionTime": 97200,
  "workspaceId": "workspace-123"
}

Privacy

COR-Matrix collects:

✅ File paths and names
✅ Line counts (additions, deletions, modifications)
✅ Timestamps
✅ Task IDs
✅ User IDs (hashed)
❌ Actual code content (NOT collected)
❌ Sensitive data (NOT collected)

Analyzing Results

Retention Metrics

High Retention (Good)

AI-generated code remains unchanged for extended periods
Indicates high-quality, production-ready output
Less rework needed

Low Retention (Needs Attention)

Frequent modifications to AI-generated code
May indicate prompt engineering improvements needed
Could suggest expert guidelines need refinement

Example Insights

Strong Performance:

File: src/utils/validation.ts
Retention: 98% (unchanged for 30 days)
Implication: AI successfully generated production-ready utility functions

Needs Improvement:

File: src/components/ComplexForm.tsx
Retention: 45% (heavily modified within 2 days)
Implication: Complex UI components may need better expert guidelines or examples

Dashboard Access

Access your COR-Matrix analytics:

Navigate to your COR-Matrix instance
Log in with your credentials
Select your workspace
View dashboards:
- Retention Overview: High-level metrics
- File Analysis: Per-file retention rates
- Team Performance: User-level insights
- Model Comparison: AI model effectiveness

Key Metrics

Overall Retention Rate: % of AI code unchanged after N days
Average Modification Time: How quickly AI code is edited
Top Retained Files: Most stable AI contributions
Refactor Hotspots: Frequently modified files
AI vs Human Code Ratio: Codebase composition

Use Cases

1. Quality Assurance

Identify which types of code generation work well:

High Retention:
- Utility functions (95%)
- Type definitions (92%)
- Test cases (88%)

Low Retention:
- Complex business logic (52%)
- UI components (63%)
- Database queries (58%)

Action: Focus AI on high-retention tasks, provide better guidance for low-retention areas. Use retention data to improve custom experts:

Observation: React components have 60% retention
Analysis: Missing accessibility guidelines
Action: Update React expert with a11y best practices
Result: Retention improved to 85%

3. Team Adoption

Track how teams use AI assistance:

Team A: 75% of commits include AI-generated code (high adoption)
Team B: 30% of commits include AI-generated code (low adoption)

Retention rates similar across both teams
→ AI quality is consistent, adoption varies
→ Provide Team B with training/examples

4. Model Comparison

Compare effectiveness of different AI models:

Model: claude-sonnet-4-5
Retention: 82%
Average Mod Time: 5.2 days

Model: gpt-4o
Retention: 78%
Average Mod Time: 3.8 days

→ Claude generates slightly more stable code
→ GPT-4o code is modified faster (may indicate faster iteration)

5. Continuous Improvement

Iterative enhancement workflow:

Generate code with HAI Build
Monitor retention in COR-Matrix
Identify low-retention patterns
Update expert guidelines
Re-generate similar code
Measure retention improvement
Repeat

Integration with CI/CD

Pre-Commit Hooks

Track retention before commits:

#!/bin/bash
# .git/hooks/pre-commit

# Send modified AI-generated files to COR-Matrix
curl -X POST https://cormatrix.example.com/api/events \
  -H "Authorization: Bearer $CORMATRIX_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "eventType": "pre_commit_check",
    "files": '$(git diff --cached --name-only | jq -R -s -c 'split("\n")[:-1]')'
  }'

GitHub Actions

Automate tracking in CI:

name: COR-Matrix Tracking

on: [push, pull_request]

jobs:
  track:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      
      - name: Report to COR-Matrix
        env:
          CORMATRIX_TOKEN: ${{ secrets.CORMATRIX_TOKEN }}
        run: |
          curl -X POST ${{ secrets.CORMATRIX_URL }}/api/events \
            -H "Authorization: Bearer $CORMATRIX_TOKEN" \
            -H "Content-Type: application/json" \
            -d '{
              "eventType": "ci_build",
              "commit": "${{ github.sha }}",
              "branch": "${{ github.ref }}",
              "workspaceId": "${{ secrets.CORMATRIX_WORKSPACE }}"
            }'

Advanced Configuration

Custom Event Types

Extend tracking with custom events:

import { corMatrixClient } from '@/services/cormatrix'

async function trackCustomEvent() {
  await corMatrixClient.sendEvent({
    eventType: 'expert_loaded',
    expertName: 'React Best Practices',
    taskId: 'task-abc123',
    timestamp: new Date().toISOString()
  })
}

Filtering

Exclude files from tracking:

// In HAI Build configuration
const config = {
  cormatrix: {
    excludePatterns: [
      'node_modules/**',
      '*.test.ts',
      'dist/**',
      '.git/**'
    ]
  }
}

Batching

Batch events for performance:

const events = [
  { eventType: 'code_generated', file: 'a.ts', ... },
  { eventType: 'code_generated', file: 'b.ts', ... },
  { eventType: 'code_modified', file: 'c.ts', ... }
]

await corMatrixClient.sendBatch(events)

Troubleshooting

Events Not Appearing

Check Configuration:

cat .hai.config | grep cormatrix

Verify Token:

curl -H "Authorization: Bearer $CORMATRIX_TOKEN" \
  https://cormatrix.example.com/api/health

Check Logs:

code ~/.cline/log/
# Look for COR-Matrix errors

Authentication Errors

Error: 401 Unauthorized

Solution: Regenerate token in COR-Matrix dashboard and update .hai.config

Network Issues

Error: ECONNREFUSED

Solution: Verify baseURL is correct and service is accessible

Best Practices

1. Regular Review

Schedule weekly/monthly reviews of retention metrics:

Identify trends
Spot areas for improvement
Celebrate wins

2. Team Training

Use retention data to train team:

Share high-retention examples
Discuss low-retention patterns
Refine prompting strategies

Continuously improve:

Generate code
Measure retention
Identify gaps
Update experts/prompts
Repeat

4. Privacy First

Never send:

Proprietary code
Customer data
Credentials
PII

Only metadata (file paths, line counts, timestamps).

5. Version Control

Track config changes:

git add .hai.config.example
git commit -m "Add COR-Matrix config template"

Provide a template without secrets:

# .hai.config.example
cormatrix.baseURL=https://your-instance.com
cormatrix.token=YOUR_TOKEN_HERE
cormatrix.workspaceId=YOUR_WORKSPACE_ID

FAQ

What data is sent to COR-Matrix?

Only metadata: file paths, line counts, timestamps, task IDs. No actual code content is transmitted.

Can I self-host COR-Matrix?

Yes, COR-Matrix can be deployed on-premises. Contact your organization’s platform team for setup.

How long is data retained?

Configurable per instance. Default is 1 year of retention history.

Does this slow down HAI Build?

No. Events are sent asynchronously and don’t block code generation.

Can I disable COR-Matrix?

Yes. Simply remove or comment out the cormatrix.* configuration in .hai.config.

​COR-Matrix Integration

​Overview

​How It Works

​Code Origin Tagging

​Retention Tracking

​Analytics Dashboard

​Configuration

​Setup via .hai.config

​Configuration Parameters

​Example Configuration

​Security Considerations

​Protecting Sensitive Data

​Environment-Specific Configuration

​Token Management

​Data Collected

​Code Change Events

​Modification Events

​Privacy

​Analyzing Results

​Retention Metrics

​Example Insights

​Dashboard Access

​Key Metrics

​Use Cases

​1. Quality Assurance

​2. Expert Refinement

​3. Team Adoption

​4. Model Comparison

​5. Continuous Improvement

​Integration with CI/CD

​Pre-Commit Hooks

​GitHub Actions

​Advanced Configuration

​Custom Event Types

​Filtering

​Batching

​Troubleshooting

​Events Not Appearing

​Authentication Errors

​Network Issues

​Best Practices

​1. Regular Review

​2. Team Training

​3. Iterative Refinement

​4. Privacy First

​5. Version Control

​FAQ

​What data is sent to COR-Matrix?

​Can I self-host COR-Matrix?

​How long is data retained?

​Does this slow down HAI Build?

​Can I disable COR-Matrix?

​Next Steps

Custom Experts

CLI Usage

COR-Matrix Integration

Overview

How It Works

Code Origin Tagging

Retention Tracking

Analytics Dashboard

Configuration

Setup via .hai.config

Configuration Parameters

Example Configuration

Security Considerations

Protecting Sensitive Data

Environment-Specific Configuration

Token Management

Data Collected

Code Change Events

Modification Events

Privacy

Analyzing Results

Retention Metrics

Example Insights

Dashboard Access

Key Metrics

Use Cases

1. Quality Assurance

2. Expert Refinement

3. Team Adoption

4. Model Comparison

5. Continuous Improvement

Integration with CI/CD

Pre-Commit Hooks

GitHub Actions

Advanced Configuration

Custom Event Types

Filtering

Batching

Troubleshooting

Events Not Appearing

Authentication Errors

Network Issues

Best Practices

1. Regular Review

2. Team Training

3. Iterative Refinement

4. Privacy First

5. Version Control

FAQ

What data is sent to COR-Matrix?

Can I self-host COR-Matrix?

How long is data retained?

Does this slow down HAI Build?

Can I disable COR-Matrix?

Next Steps