Skip to content

AI_INTEGRATION_GUIDE.md

Latest commit

 

History

History
239 lines (182 loc) · 5.23 KB

AI_INTEGRATION_GUIDE.md

File metadata and controls

239 lines (182 loc) · 5.23 KB

TaskTracker AI Integration Guide

Overview

TaskTracker is designed to be the memory layer between you and your AI assistants. This guide shows how to effectively use TaskTracker with various AI tools.

Quick Start for AI Sessions

1. Starting a New Session

# Get immediate context
tt c

# Or for full context
tt cf

Copy the output and paste it at the beginning of your AI conversation. This gives the AI:

  • Your current work status
  • Recent decisions
  • Active blockers
  • Project goals

2. During Development

# Document progress
tt j "Implemented user authentication with JWT tokens"

# Record decisions
tt j --type decision "Using PostgreSQL for better JSON support"

# Note blockers
tt j --type blocker "CORS issues with frontend API calls"

# Tag important work
tt j --tags auth,security "Added password hashing with bcrypt"

3. Switching AI Tools

When moving from one AI to another (e.g., Cursor to ChatGPT):

# Generate fresh context
tt cf --output context.md

# Then paste the content into your new AI session

Integration Patterns

Pattern 1: Continuous Context (Recommended)

# Start of session
tt c

# After significant progress
tt j "Completed user registration flow"
tt c  # Regenerate context

# Before complex questions
tt cf  # Get full context for AI

Pattern 2: Session-Based Development

# Morning startup
tt cf --output morning-context.md

# Throughout the day
tt j "progress updates..."

# End of day
tt j --type decision "Tomorrow: focus on API optimization"
tt journal-export markdown --output today-journal.md

Pattern 3: Feature-Focused Context

# Tag all work on a feature
tt j --tags auth "Starting OAuth implementation"
tt j --tags auth "Added Google OAuth provider"

# Get feature-specific context
tt journal-show --tag auth

AI Tool-Specific Tips

Cursor

  1. Use tt c output in your system prompt
  2. Update context after file changes
  3. Use tt j --files filename.js to track file-specific work

ChatGPT/Claude

  1. Start conversations with tt cf output
  2. Use tt journal-export for long conversations
  3. Reference specific decisions: tt journal-search "architecture"

GitHub Copilot

  1. Keep context.md in your project root
  2. Update it regularly: tt cf --output context.md
  3. Reference it in comments for better suggestions

Advanced Workflows

1. PRD-Driven Development

# Set project vision
tt prd requirements.md

# Generate PRD context for AI
tt prd-context

# Combine with journal
tt cf  # Includes both PRD and journal

2. Debugging with Context

# Document the issue
tt j --type blocker "API returns 500 on user creation"

# Add relevant files
tt j --files "api/users.js,models/user.js" "Investigating user creation bug"

# Generate debug context
tt cf 1  # Last day only for focused context

3. Architecture Decisions

# Document major decisions
tt j --type decision --tags architecture "Microservices for scalability"

# Export architecture decisions
tt journal-search "architecture" 
tt journal-export markdown --output architecture-decisions.md

Best Practices

1. Context Hygiene

  • Regenerate context at the start of each session
  • Use tt c for quick questions
  • Use tt cf for complex implementation tasks

2. Meaningful Entries

❌ Bad: tt j "fixed bug" ✅ Good: tt j "Fixed null pointer in user auth when email is missing"

3. Decision Documentation

Always document "why" not just "what":

tt j --type decision "Chose Redis for session storage due to auto-expiration feature"

4. Blocker Patterns

When stuck, document thoroughly:

tt j --type blocker "CORS error: 'Access-Control-Allow-Origin' missing"
tt j "Tried adding cors middleware - still failing"
tt j --type decision "Solution: Configure CORS before other middleware"

JSON API for Tool Integration

TaskTracker supports JSON output for programmatic use:

# Get journal entries as JSON
tt journal-show --json

# Search with JSON output
tt journal-search "auth" --json

# Export for processing
tt journal-export json --output data.json

JSON Structure

{
  "success": true,
  "data": {
    "entries": [{
      "id": 1234567890,
      "timestamp": "2024-01-15T10:30:00Z",
      "type": "progress",
      "content": "Implemented user authentication",
      "tags": ["auth", "backend"],
      "files": ["auth.js"],
      "session": "2024-1-15-10"
    }]
  }
}

Troubleshooting

Context Too Long

# Limit to recent days
tt cf 3  # Last 3 days only

# Or use quick context
tt c  # Essential info only

Missing Context

# Check journal entries
tt js

# Verify PRD
tt prd-show

# Regenerate everything
tt cf --output full-context.md

Stale Context

# Always regenerate after breaks
tt c  # Quick refresh

# Document session boundaries
tt j --type context "Resuming after 2-day break"

Future Integrations

TaskTracker is designed to be extended. Planned integrations:

  • VS Code extension for automatic context updates
  • GitHub Actions for PR context
  • Slack bot for team context sharing
  • API endpoints for custom integrations

Remember: TaskTracker is your development memory. The more you document, the better your AI assistants can help you.