polymind-cli

v1.0.0

Published

7 months ago

PolyMind CLI - Multi-model AI council debates in your terminal with interactive TUI

Downloads

0High
0Medium
0Low

sinisterchill

ai cli gemini debate council terminal tui llm multi-agent chat interactive

PolyMind CLI

The LLM Council: Let multiple AI models collaborate, review each other's work, and synthesize the best response

🏛️ The LLM Council

Instead of querying a single AI model, PolyMind creates a council of AI models that work together:

Stage 1 - First Opinions: All models respond to your query individually
Stage 2 - Peer Review: Models review and rank each other's responses (anonymized)
Stage 3 - Chairman Synthesis: A designated Chairman compiles the final answer

Why use an LLM Council?

✅ Eliminates single-model bias - Multiple perspectives reduce errors
✅ Peer review improves quality - Models critique each other anonymously
✅ Best-of-breed synthesis - Chairman combines the strongest insights
✅ Transparency - See all individual responses in a tab view

🌟 What's New

v1.0.0 brings the LLM Council system with:

🏛️ LLM Council Mode - Multi-model collaboration with 3-stage workflow
🔥 Live Streaming - Real-time AI responses with token-by-token streaming
📊 Peer Review System - Models rank each other anonymously
🎯 Chairman Synthesis - Final response from the council's best insights
📚 RAG Support - Context-aware responses from your codebase
🔌 Multi-Provider - Gemini, GPT, Claude, Grok, and more
🎨 Beautiful TUI - Tab view for comparing model responses

🚀 Quick Start

# Install globally
npm install -g @polymind/cli

# Configure your API keys (at least 2 providers recommended)
polymind init
export OPENAI_API_KEY="your-key"
export ANTHROPIC_API_KEY="your-key"

# Query the LLM Council
polymind council "What is the future of artificial intelligence?"

# Or start live chat with a single model
polymind live --rag

✨ Key Features

🏛️ LLM Council Mode (NEW!)

Harness the collective intelligence of multiple AI models working together.

polymind council "Explain quantum computing"

How it works:

Stage 1: First Opinions

Query sent to all council members (GPT, Gemini, Claude, Grok)
Each model responds independently
View responses in an interactive tab interface

Stage 2: Peer Review

Each model reviews OTHER models' responses (anonymized as "Response A", "Response B")
Models rank responses by accuracy and insight
Prevents favoritism and bias

Stage 3: Chairman Synthesis

Designated Chairman reviews all responses and rankings
Synthesizes the best insights into one final answer
Shows which sources were incorporated

Example Council:

polymind council "Should we use microservices?" \
  --gemini-key $GEMINI_API_KEY \
  --openai-key $OPENAI_API_KEY \
  --anthropic-key $ANTHROPIC_API_KEY \
  --chairman "Claude Sonnet"

Gemini Deep Reasoning Preset:

# Use the highest reasoning Gemini models
polymind council "Explain quantum entanglement" --gemini-deep

# With custom chairman
polymind council "Your question" --gemini-deep --chairman "Gemini 2.5 Pro"

This preset includes:

Chairman: Gemini 3 Pro Preview
Members: Gemini 3 Pro Preview, Gemini 2.5 Pro, Gemini 2.5 Flash

🔴 Live Streaming Mode

Experience real-time AI responses with token-by-token streaming, just like ChatGPT but in your terminal.

polymind              # Launch live chat
polymind live --rag   # Enable RAG for context

Features:

Real-time streaming responses
Automatic codebase indexing
Slash command support (/help, /rag, /model, /exit)
Context-aware conversations
Model switching on the fly

📚 RAG (Retrieval-Augmented Generation)

Automatically indexes your workspace to provide context-aware responses about your code.

polymind live --rag  # Enable RAG mode

What it does:

Indexes .ts, .tsx, .js, .jsx, .md, .json files
Semantic search across your codebase
Injects relevant context into AI queries
Real-time stats: 📚 Indexed 247 files (1.2MB)

🎭 Live Multi-Agent Debates

Watch multiple AI personas deliberate on complex topics in real-time.

polymind debate "Should we colonize Mars?"
polymind debate "Is AGI achievable by 2030?" --rounds 4

Council Members:

🎯 Pragmatist - Practical problem solver
🚀 Visionary - Future-focused innovator
🤔 Skeptic - Critical analyst
⚖️ Ethicist - Moral compass

Each persona streams their argument in real-time, then a synthesized conclusion is generated.

🔌 Multi-Model Support

Switch between providers and models dynamically:

# Configure your provider
polymind init --provider gemini

# Switch models mid-conversation
/model gemini-2.0-flash-exp
/model gpt-4-turbo
/model claude-3-opus

Supported Providers:

Google Gemini - gemini-2.0-flash-exp, gemini-1.5-pro, gemini-1.5-flash
OpenAI - gpt-4-turbo, gpt-4, gpt-3.5-turbo (coming soon)
Anthropic - claude-3-opus, claude-3-sonnet, claude-3-haiku (coming soon)

📚 Commands

Interactive Commands

| Command | Alias | Description | |---------|-------|-------------| | polymind | - | Launch live streaming chat (default) | | polymind live | l | Explicitly start live mode | | polymind debate <question> | - | Start multi-agent council debate | | polymind init | - | Configure API keys and settings | | polymind status | - | Check configuration |

Slash Commands (In Chat)

| Command | Description | |---------|-------------| | /help | Show available commands | | /rag | Toggle RAG mode | | /model <name> | Switch AI model | | /clear | Clear chat history | | /exit | Quit application | | /debate <topic> | Start debate mode |

🔧 Configuration

Initial Setup

polymind init

You'll be prompted for:

AI provider (Gemini, OpenAI, Anthropic)
API key
Default model
Feature preferences (streaming, RAG, etc.)

Manual Configuration

Config is stored in ~/.config/polymind/config.json:

{
  "provider": "gemini",
  "apiKey": "your-api-key-here",
  "streaming": true,
  "quirks": true,
  "animations": true
}

Environment Variables

export POLYMIND_API_KEY="your-key"
export POLYMIND_PROVIDER="gemini"

🎨 Advanced Usage

Live Chat with RAG

# Enable RAG to query your codebase
polymind live --rag

# In chat:
> Explain how the streaming client works
> What files implement the debate logic?
> Summarize the RAG engine architecture

Custom Debate Parameters

# More rounds
polymind debate "Future of work" --rounds 5

# Specific model
polymind debate "AI ethics" --model gemini-1.5-pro --rounds 3

Keyboard Shortcuts

ESC - Exit application
Ctrl+C (during stream) - Cancel current response
Ctrl+C (idle) - Exit application

🏗️ Architecture

Inspired by Gemini CLI, PolyMind uses:

Ink - React for terminal UIs
Google Generative AI SDK - For Gemini models
Streaming Architecture - Real-time token-by-token responses
RAG Engine - Document indexing and semantic search
Multi-Agent System - Parallel persona processing

📦 Installation

From npm

npm install -g @polymind/cli

From Source

git clone https://github.com/reyyanxahmed/polymind-cli.git
cd polymind-cli
npm install
npm run build
npm link

🚀 Deployment

For Developers

# Build and test locally
npm run build
npm link

# Deploy to npm
./scripts/deploy.sh 1.0.0

CI/CD

GitHub Actions automatically publishes to npm when you push a version tag:

git tag v1.0.0
git push origin v1.0.0

git push origin v1.0.0


## 🛠️ Development

### Local Development

```bash
# Clone the repo
git clone https://github.com/reyyanxahmed/polymind-cli.git
cd polymind-cli

# Install dependencies
npm install

# Run in development mode
npm run dev

# Build for production
npm run build

# Link globally for testing
npm link

Project Structure

polymind-cli/
├── src/
│   ├── core/
│   │   ├── streaming-client.ts    # Multi-model streaming client
│   │   └── rag-engine.ts          # RAG indexing and search
│   ├── commands/
│   │   ├── live.ts                # Live streaming chat command
│   │   ├── live-debate.ts         # Live debate command
│   │   └── init.ts                # Setup wizard
│   ├── ui/
│   │   ├── apps/
│   │   │   ├── LiveChatApp.tsx    # Main streaming chat UI
│   │   │   └── LiveDebateApp.tsx  # Multi-agent debate UI
│   │   └── components/
│   └── utils/
│       └── slash-commands.ts      # Command parsing
├── package.json
└── README.md

🤝 Contributing

We welcome contributions! Areas where you can help:

Provider Integrations - OpenAI, Anthropic, Ollama, etc.
RAG Enhancements - Better embeddings, vector databases
UI/UX - More animations, themes, customization
Features - Image support, voice, web search
Documentation - Tutorials, examples, translations

See CONTRIBUTING.md for guidelines.

📄 License

🔗 Links

GitHub: https://github.com/reyyanxahmed/polymind-cli
npm: https://www.npmjs.com/package/@polymind/cli
Issues: https://github.com/reyyanxahmed/polymind-cli/issues
Inspiration: Gemini CLI

🙏 Credits

Built with inspiration from:

Gemini CLI - Architecture and streaming design
Ink - React for terminal UIs
Commander.js - CLI framework

Made with ❤️ by the PolyMind Team

Type polymind and let the AI council begin!