Catcode 🐱

A sentiment-aware ASCII cat companion for Claude Code's statusline that generates dynamic motivational phrases.

catcode.mov

What It Does

Catcode analyzes your conversation with Claude and displays a cute ASCII cat with contextual phrases in your Claude Code statusline:

(=^.^=) | Stay focused today
(=o.o=) | Deep thinking mode  
(=-.=) | Progress through struggle

The cat faces and phrases are dynamically generated by Claude's Haiku model based on the sentiment of your latest message - nothing is hardcoded!

Features

• Dual Mode Operation: Works as both a standalone console tool and Claude Code statusline integration
• Dynamic Generation: Cat ASCII and motivational phrases are generated live by Claude Haiku model
• Sentiment Analysis: Responds appropriately to positive, negative, and neutral messages
• Smart Caching: UUID-based caching prevents API spam during statusline updates
• Fast Performance: Sub-second cached responses suitable for ~300ms statusline updates

Requirements

• Node.js 18+
• Claude Code installed: npm install -g @anthropic-ai/claude-code
• Authenticated Claude CLI: claude doctor

Quick Setup

1. Clone and setup:

   git clone <this-repo>
   cd catcode
   chmod +x catcode.sh

2. Test console mode:

   ./catcode.sh "I'm feeling great today!"
   # Output: (=^.^=) | Celebrate small victories

3. Configure statusline in ~/.claude/settings.json:

   {
     "statusLine": {
       "type": "command", 
       "command": "/path/to/catcode/catcode.sh",
       "padding": 0
     }
   }

Usage Examples

Console Mode:

# Direct argument
./catcode.sh "This is frustrating"
# Output: (=-.=) | Patience builds strength

# Via stdin  
echo "Working on the project" | ./catcode.sh
# Output: (=o.o=) | Steady progress counts

Statusline Mode: When configured in Claude Code settings, catcode automatically:

• Receives session JSON via stdin
• Parses the transcript file to find your latest message
• Generates contextual cat + phrase based on sentiment
• Updates your statusline with motivational companion

How It Works

1. Input Processing: Detects JSON (statusline mode) vs plain text (console mode)
2. Message Extraction: For statusline mode, parses Claude Code JSONL transcript to get latest user message
3. Sentiment Analysis: Uses Claude Haiku to classify sentiment (positive/negative/neutral)
4. Dynamic Generation: Prompts Claude to generate appropriate cat ASCII + 3-5 word motivational phrase
5. Caching: Stores results by message UUID to avoid redundant API calls

Project Status

✅ PRODUCTION READY - Full dual-mode operation implemented
✅ Phase 1 COMPLETE - Console prototype with dynamic generation
✅ Phase 2 COMPLETE - Statusline integration with real JSONL parsing

See roadmap.md for detailed development progress.

Technical Details

• Model: claude-3-5-haiku-20241022 for fast, cost-effective generation
• Cache Location: ~/.cache/catcode/ (respects XDG_CACHE_HOME)
• Response Format: (<cat-ascii>) | <motivational-phrase>
• Update Frequency: Optimized for Claude Code's ~300ms statusline refresh rate

Troubleshooting

• No output: Check claude doctor and model availability
• Wrong format: Verify jq is installed for JSON parsing
• Slow responses: Check cache directory permissions
• API errors: Ensure Claude CLI is authenticated

---

A fun little project demonstrating Claude Code's statusline feature with headless API calls and transcript parsing.