Claude-Code-Workflow/.claude/scripts

pycli - Python CLI Wrapper with Hierarchical Vector Database

This directory contains the bash wrapper and configuration for the enhanced Python-based analysis CLI with hierarchical vector database support.

📁 Files

• pycli - Main bash wrapper script
• pycli.conf - Configuration file
• install_pycli.sh - Installation script
• README.md - This documentation

🚀 Quick Installation

# Run the installation script
bash install_pycli.sh

# Follow the prompts to configure your shell
# The script will automatically detect your Python installation

# Verify installation
pycli --help

🎯 Key Features

Hierarchical Vector Database

• Smart Parent Discovery: Subdirectories automatically use parent's vector database
• No Redundant Processing: Avoids duplicate vectorization in project subdirectories
• Central Storage: All vector databases stored in ~/.claude/vector_db/
• Path-based Organization: Organized by project directory structure

Unified Interface

• Single Command: pycli replaces complex Python script calls
• Intelligent Context: Automatic file discovery with semantic search
• Tool Integration: Seamless integration with Gemini and Codex
• Configuration Management: Environment-specific Python interpreter paths

📋 Common Commands

# Initialize new project
cd /path/to/your/project
pycli --init

# Smart analysis
pycli --analyze --query "authentication patterns" --tool gemini

# Direct analysis
pycli --analyze --tool codex -p "implement user login"

# Maintenance
pycli --update-embeddings
pycli --status

🔧 Configuration

Edit ~/.claude/scripts/pycli.conf after installation:

# Python interpreter path
PYTHON_PATH="/usr/bin/python3"

# Vector database root directory
VECTOR_DB_ROOT="$HOME/.claude/vector_db"

# Python scripts directory
PYTHON_SCRIPT_DIR="$HOME/.claude/python_script"

🏗️ How Hierarchical DB Works

Project Structure:          Vector Database:
/home/user/myproject/       ~/.claude/vector_db/
├── src/                    └── home_user_myproject/
│   ├── auth/                   ├── embeddings.pkl
│   └── api/                    └── index.json
└── tests/

# All subdirectories use the single parent DB

📖 Documentation

For complete usage information, see:

• Strategy Guide: ~/.claude/workflows/python-tools-strategy.md
• Installation Guide: Run bash install_pycli.sh for guided setup

🎪 Migration from Legacy Tools

# Replace gemini-wrapper
# OLD: ~/.claude/scripts/gemini-wrapper -p "prompt"
# NEW: pycli --analyze --tool gemini -p "prompt"

# Replace codex commands
# OLD: codex --full-auto exec "task"
# NEW: pycli --analyze --tool codex -p "task"

# Enhanced with context discovery
pycli --analyze --query "relevant context" --tool both

🐛 Troubleshooting

# Check system status
pycli --status

# Rebuild everything
pycli --init

# Test search functionality
pycli --test-search

# View configuration
cat ~/.claude/scripts/pycli.conf

💡 Advanced Usage

Project Integration

# Add to package.json
{
  "scripts": {
    "analyze": "pycli --analyze --query",
    "ai-init": "pycli --init",
    "ai-update": "pycli --update-embeddings"
  }
}

# Use in Makefiles
analyze:
	pycli --analyze --query "$(QUERY)" --tool gemini

CI/CD Integration

# GitHub Actions example
- name: Update AI Context
  run: pycli --update-embeddings

- name: Analyze Changes
  run: pycli --analyze --query "code review" --tool gemini

---

For questions or issues, check the documentation or run pycli --help.