YesMan 🤖

Smart terminal wrapper with multi-provider LLM automation

YesMan is a TUI (Terminal User Interface) application that monitors CLI program output and automatically handles interactive prompts using Large Language Models. It shows your program's output in a terminal-in-terminal view while intelligently responding to questions and prompts.

🤔 Why Do You Need This?

YesMan enables fully automated AI pair programming with visual oversight.

When using AI coding tools like Codex CLI or Claude Code, you constantly face interrupting prompts:

• "Apply this change? [Y/n]"
• "Create new file? [Y/n]"
• "Run this command? [Y/n]"
• "Continue with refactoring? [Y/n]"
• "Install dependencies? [Y/n]"

The Problem: These safety prompts interrupt the AI's flow, requiring manual intervention every few seconds and defeating the purpose of automation.

The Solution: YesMan acts as an intelligent supervisor that:

✅ Auto-Approves AI Suggestions

• Let Claude Code refactor entire codebases while you watch
• Enable true "autopilot" coding sessions where AI completes full features
• Remove friction from AI-driven development workflows

👀 Provides Full Visual Control

• See exactly what the AI is doing in real-time
• Pause or intervene at any moment with a single keypress
• Terminal-in-terminal view shows all AI actions transparently

🚀 Enables Autonomous Development

• Run overnight AI-driven development sessions
• Automate repetitive coding tasks with AI assistance
• Build proof-of-concepts with minimal human intervention

🛡️ Maintains Safety

• Smart context-aware defaults for AI tool prompts
• Manual override capability always available
• Pause countdown before executing any action

💡 Perfect For AI Coding Workflows

• Feature Development: Let AI implement entire features while you supervise
• Codebase Refactoring: Automated large-scale code improvements
• Dependency Management: AI handles package installations and updates
• Testing & CI: Automated test writing and pipeline setup
• Documentation: AI generates docs while handling all confirmations
• Code Reviews: AI applies suggested changes across multiple files

✨ Features

• 🖥️ Terminal-in-terminal display using PTY (pseudo-terminal)
• 🔍 Real-time output monitoring with intelligent pattern detection
• 🤖 Multi-provider LLM support (Ollama, OpenAI, Anthropic, Groq)
• 🎮 Interactive controls with pause, manual mode, and help
• ⚡ Response caching for faster repeated interactions
• 🎯 Smart prompt detection for various question formats
• 🔧 Manual intervention available at any time
• 📊 Rich TUI interface with status, diagnostics, and provider info

🚀 Quick Start

Installation

# Clone or download yesman.py (single file solution)
chmod +x yesman.py

# Install dependencies
pip install rich requests

Setup LLM Provider

Option 1: Ollama (Recommended - Local & Fast)

# Install Ollama
curl -fsSL https://ollama.ai/install.sh | sh

# Pull a fast model
ollama pull qwen3:4b

# YesMan will auto-detect Ollama
./yesman.py python your_script.py

Option 2: OpenAI

export OPENAI_API_KEY="your-api-key"
./yesman.py --provider openai python your_script.py

Option 3: Other Providers

# Anthropic Claude
export ANTHROPIC_API_KEY="your-api-key"
./yesman.py --provider anthropic python your_script.py

# Groq (fast cloud inference)
export GROQ_API_KEY="your-api-key"
./yesman.py --provider groq python your_script.py

📖 Usage

Basic Usage

# Auto-detect best provider
./yesman.py python install_script.py

# Use specific provider and model
./yesman.py --provider ollama --model qwen3:4b ./configure.sh

# Manual mode (no automation)
./yesman.py --manual npm install

# Full auto mode (no pause)
./yesman.py --auto apt update

Interactive Controls

• Space - Pause automation during countdown
• Enter - Accept LLM suggestion (when paused)
• Esc - Switch to manual mode (always available)
• a - Enable auto mode
• m - Switch to manual mode
• r - Retry LLM on error
• ? - Toggle help
• p - Show provider info
• d - Show diagnostics

AI Coding Examples

# Let Claude Code refactor entire codebase autonomously
./yesman.py claude-code "refactor this React app to use TypeScript"

# Automated feature development with oversight
./yesman.py codex "implement user authentication system"

# AI-driven dependency management
./yesman.py claude-code "update all packages and fix breaking changes"

# Autonomous testing and CI setup
./yesman.py --auto claude-code "add comprehensive tests for all components"

# Full application scaffolding
./yesman.py codex "create a complete CRUD API with database setup"

🔧 How It Works

1. PTY Subprocess: Spawns your command in a pseudo-terminal
2. Pattern Detection: Monitors output for questions using regex patterns
3. LLM Analysis: Sends context to LLM when prompts are detected
4. Smart Response: LLM suggests appropriate responses (y/n/enter/etc.)
5. Auto-execution: Sends response after countdown (with pause option)

Supported Prompt Patterns

• [Y/n], [y/N] - Yes/no questions
• (yes/no) - Confirmation prompts
• Press ENTER - Continue prompts
• ? - General questions
• Continue?, Proceed? - Confirmation
• Choose 1/2/3 - Selection menus

🤖 LLM Providers

Provider	Models	Setup	Speed	Cost
Ollama	qwen3:4b, phi4-mini, devstral	Local install	⚡ Fast	🆓 Free
OpenAI	gpt-3.5-turbo, gpt-4o-mini, gpt-4o	API key	🔥 Very Fast	💰 Paid
Anthropic	claude-3-haiku, claude-3.5-sonnet	API key	🚀 Fast	💰 Paid
Groq	llama-3.1-8b-instant, llama3-70b	API key	⚡ Very Fast	🆓 Free tier

Model Recommendations

• Speed: qwen3:4b (Ollama), gpt-3.5-turbo (OpenAI)
• Balance: phi4-mini (Ollama), gpt-4o-mini (OpenAI)
• Quality: devstral (Ollama), gpt-4o (OpenAI)

📋 Command Line Options

./yesman.py [OPTIONS] COMMAND [ARGS...]

Options:
  --provider {auto,ollama,openai,anthropic,groq}  LLM provider
  --model MODEL                                   Specific model to use
  --auto                                          Full auto mode (no pause)
  --manual                                        Manual mode (no automation)
  --pause SECONDS                                 Pause duration (default: 3)
  --cache-file FILE                               Response cache file
  --list-providers                                Show available providers
  --setup-ollama                                  Setup Ollama with models

🛠️ Advanced Configuration

Environment Variables

# Ollama configuration
export OLLAMA_HOST="http://localhost:11434"  # Custom Ollama server

# API Keys
export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="sk-ant-..."  
export GROQ_API_KEY="gsk_..."

Response Caching

YesMan caches LLM responses in .yesman_cache.json to speed up repeated interactions with the same prompts.

Remote Ollama

# Use remote Ollama instance
export OLLAMA_HOST="http://192.168.1.100:11434"
./yesman.py python script.py

🧪 Testing

Test YesMan with the included interactive test application:

# Basic test
./yesman.py python test_interactive_app.py

# Test with specific provider
./yesman.py --provider ollama --model qwen3:4b python test_interactive_app.py

The test app simulates various prompt types to validate automation capabilities.

🔍 Troubleshooting

Check Available Providers

./yesman.py --list-providers echo test

Ollama Issues

# Check Ollama status
./yesman.py --setup-ollama

# Pull missing models
ollama pull qwen3:4b

# Check running models
ollama ps

Debug Mode

# Enable diagnostics view
# Press 'd' during execution to see provider diagnostics

🤝 Contributing

YesMan is a single-file solution for easy deployment and modification. Feel free to:

• Add new LLM providers
• Improve prompt detection patterns
• Enhance the TUI interface
• Add new automation features

📄 License

MIT License - see LICENSE file for details.

🙏 Acknowledgments

• Built with Rich for beautiful TUI
• LLM integration supports multiple providers
• PTY handling for true terminal emulation