🗣️ Multi-LLM Group Chat Discussion

What happens when 5 different AI models discuss the same topic?

Simulate a human-like group chat discussion using 5 independent Large Language Models to analyze any topic, question, or document. Each LLM has a distinct thinking style, creating diverse perspectives that surface insights no single model would find alone.

✨ What It Does

Think of it as five friends with different thinking styles discussing something in a group chat:

• 🧠 Logical Thinker - Structured, fact-focused, breaks things down
• 🎨 Creative Thinker - Imaginative, speculative, alternative viewpoints
• 🤔 Skeptical Thinker - Questions assumptions, plays devil's advocate
• 🔧 Practical Thinker - Real-world feasibility, cost, risks
• ⚖️ Synthesizer - Observes and bridges different viewpoints

🎯 Use Cases

• Exploring complex decisions from multiple angles
• Stress-testing ideas before presenting them
• Finding blind spots in your thinking
• Research and analysis with built-in devil's advocate
• Brainstorming with diverse AI perspectives

🚀 Quick Start

1. Clone the repo

git clone https://github.com/JeetInTech/Multi-LLM-Discussion.git
cd Multi-LLM-Discussion

2. Install dependencies

pip install -r requirements.txt

3. Set up your API keys

# Copy the example env file
cp .env.example .env

# Edit .env and add your Groq API key (free)
# Get one at: https://console.groq.com/keys

4. Install Ollama (for local models)

Download from ollama.com, then:

ollama pull llama3.2
ollama pull mistral
ollama pull phi3

5. Run a discussion

python main.py "Should AI be regulated?"

🔧 Configuration

The system uses 5 different models for true diversity:

Persona	Provider	Model
🧠 Logical	Ollama (local)	llama3.2
🤔 Skeptical	Ollama (local)	mistral
⚖️ Synthesizer	Ollama (local)	phi3
🎨 Creative	Groq (cloud)	llama-3.3-70b-versatile
🔧 Practical	Groq (cloud)	qwen/qwen3-32b

All Free LLM Options

Provider	Cost	Setup
Ollama	100% Free	Runs locally on your GPU
Groq	Free tier	Get API key
Google Gemini	Free tier	Get API key
HuggingFace	Free tier	Get API key

📁 Project Structure

Multi-LLM-Discussion/
├── main.py           # CLI entry point
├── discussion.py     # Core discussion engine
├── personas.py       # 5 persona definitions
├── llm_clients.py    # API clients (Ollama, Groq, etc.)
├── config.py         # Configuration settings
├── demo_offline.py   # Demo without API keys
├── .env.example      # Template for API keys
├── requirements.txt  # Python dependencies
└── README.md

💻 Command Line Options

python main.py [topic] [options]

Options:
  --file, -f FILE    Read topic from a file
  --rounds, -r N     Number of discussion rounds (default: 3)
  --no-synth         Disable the Neutral Synthesizer

📋 Example Output

[User]
Should remote work become the default?

─── Round 1 ───

[🧠 Logical Thinker]
Let's look at the data. Studies show remote workers are often 
more productive, with less commute stress...

[🎨 Creative Thinker]
What if the office isn't about work at all? Maybe it's about 
spontaneous collisions of ideas...

[🤔 Skeptical Thinker]
But those productivity studies mostly come from self-reporting. 
How do we actually measure creativity remotely?...

[🔧 Practical Thinker]
The real question is: can you maintain culture and onboard 
new people effectively?...

🔒 Safety Design

This is NOT an autonomous agent framework. It's a conversation simulator.

• ❌ No tool usage
• ❌ No memory persistence
• ❌ No self-reflection loops
• ❌ No planning or execution
• ✅ Each LLM responds only to visible chat history
• ✅ Pure text generation simulation

🤝 Contributing

Contributions welcome! Ideas:

• Additional personas
• New LLM providers
• Export formats (Markdown, HTML, JSON)
• Web interface

📄 License

MIT License - Free to use for any purpose. See LICENSE file.

👤 Author

JeetInTech - GitHub

---

⭐ Star this repo if you find it useful!