Multi-LLM Discussion System

A real-time discussion platform that orchestrates conversations between different Large Language Models (LLMs), enabling AI-to-AI debates with human intervention capabilities.

Features

🤖 Multi-LLM Support: GPT-5.1, GPT-4, Claude Sonnet 4.5, AWS Bedrock models, and more
☁️ Multiple Providers: OpenAI, Anthropic, and AWS Bedrock
💬 Real-time Streaming: Token-by-token conversation updates via Server-Sent Events
🎭 Role-based Discussion: Primary LLM presents ideas, Critic LLM evaluates
👤 Human Intervention: Join discussions at any time with your own messages
📋 Copy to Markdown: Export entire conversations with timestamps
🎨 Clean UI: Modern, responsive interface with expandable configurations
⚡ Performance Optimized: Token batching prevents browser freezing during streaming

Tech Stack

Frontend

React 18 + TypeScript
Vite
Tailwind CSS
Zustand (state management)
react-markdown

Backend

Node.js 20 + Express + TypeScript
OpenAI SDK
Anthropic SDK
AWS SDK for Bedrock Runtime
Server-Sent Events (SSE)

Quick Start

Prerequisites

Node.js 20.x or higher
OpenAI API key
Anthropic API key
AWS credentials (optional, only for Bedrock models)

Option 1: Using Docker (Recommended)

Clone the repository

git clone https://github.com/shaharia-lab/multi-llm-discussion.git
cd multi-llm-discussion

Set up environment variables

cp .env.example .env
# Edit .env and add your API keys

Start with Docker Compose
```
docker-compose up
```
Open your browser Navigate to http://localhost:3000

Option 2: Local Development

Clone the repository

git clone https://github.com/shaharia-lab/multi-llm-discussion.git
cd multi-llm-discussion

Install dependencies
```
pnpm install
```

Set up environment variables

cp .env.example .env
# Edit .env with your API keys:
# OPENAI_API_KEY=your_openai_key
# ANTHROPIC_API_KEY=your_anthropic_key

Start development servers
```
pnpm dev
```
Open your browser Navigate to http://localhost:3000

Environment Variables

Create a .env file in the root directory:

# Required
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...

# Optional - Only needed for AWS Bedrock models
AWS_ACCESS_KEY_ID=AKIA...
AWS_SECRET_ACCESS_KEY=...
AWS_REGION=eu-west-1

# Server Configuration
PORT=3001

Supported Models

OpenAI

GPT-5.1 (gpt-5.1-2025-11-13)
GPT-4
GPT-3.5 Turbo

Anthropic

Claude Sonnet 4.5 (claude-sonnet-4-5-20250929)
Claude 3 Opus

AWS Bedrock

Claude Sonnet 4.5 (Bedrock) (eu.anthropic.claude-sonnet-4-5-20250929-v1:0)
Claude Opus 4 (Bedrock) (eu.anthropic.claude-opus-4-20250514-v1:0)

How It Works

Configure Discussion: Set a topic and choose Primary/Critic LLM models with custom system prompts
Start Discussion: The Primary LLM presents ideas on the topic
Critic Responds: The Critic LLM evaluates and critiques the Primary's response
Back and Forth: LLMs continue the discussion automatically
Human Intervention: Jump in anytime by typing your message
Export: Copy the entire conversation as markdown for later use

Usage Guide

Starting a Discussion

Enter a Topic: Type your discussion topic in the large expandable textarea
Configure Primary LLM (Optional): Click to expand and customize
- Select model (GPT-5.1, GPT-4, etc.)
- Customize system prompt
Configure Critic LLM (Optional): Click to expand and customize
- Select model (Claude Sonnet 4.5, etc.)
- Customize system prompt
Click "Start Discussion": The discussion will begin automatically

During a Discussion

Watch the Discussion: Messages will stream in real-time with color-coded bubbles:
- 🟢 Green: OpenAI models (GPT)
- 🟠 Orange: Anthropic models (Claude) and Bedrock models
- 🟡 Yellow: Your messages
Intervene: Type a message in the input field at the bottom to join the conversation
- Your message will be sent to the Primary LLM
- The Primary will respond to you
- The Critic will then evaluate the Primary's response
- The discussion continues automatically
Stop Discussion: Click the "Stop Discussion" button to end the conversation

Discussion Flow

Primary LLM → Presents idea on topic
    ↓
Critic LLM → Evaluates and critiques
    ↓
Primary LLM → Responds to critique
    ↓
Critic LLM → Further evaluation
    ↓
(Loop continues until stopped)

Project Structure

multi-llm-discussion/
├── backend/
│   ├── src/
│   │   ├── adapters/
│   │   │   ├── openai.ts          # OpenAI API integration
│   │   │   ├── anthropic.ts       # Anthropic API integration
│   │   │   └── bedrock.ts         # AWS Bedrock API integration
│   │   ├── discussionController.ts # Discussion orchestration
│   │   ├── streamManager.ts        # SSE stream handling
│   │   ├── types.ts                # TypeScript interfaces
│   │   └── index.ts                # Express server
│   ├── package.json
│   └── tsconfig.json
├── frontend/
│   ├── src/
│   │   ├── components/
│   │   │   ├── ConfigurationForm.tsx
│   │   │   ├── DiscussionView.tsx
│   │   │   ├── MessageBubble.tsx
│   │   │   └── InterventionInput.tsx
│   │   ├── App.tsx                 # Main application
│   │   ├── store.ts                # Zustand state management
│   │   ├── types.ts                # TypeScript interfaces
│   │   └── main.tsx                # Entry point
│   ├── package.json
│   ├── vite.config.ts
│   └── tailwind.config.js
├── .env.example
├── package.json
├── pnpm-workspace.yaml
└── README.md

API Endpoints

POST `/api/discussions/start`

Start a new discussion

Request:

{
  "topic": "Discussion topic",
  "participants": [
    {
      "id": "uuid",
      "modelId": "gpt-4",
      "provider": "openai",
      "displayName": "GPT-4",
      "systemPrompt": "...",
      "role": "primary"
    },
    {
      "id": "uuid",
      "modelId": "eu.anthropic.claude-sonnet-4-5-20250929-v1:0",
      "provider": "bedrock",
      "displayName": "Claude Sonnet (Bedrock)",
      "systemPrompt": "...",
      "role": "critic"
    }
  ]
}

Response:

{
  "discussionId": "uuid"
}

GET `/api/discussions/:id/stream`

Server-Sent Events endpoint for streaming messages

Events:

token: Individual token from LLM response
complete: Message generation complete
message_start: New message started
error: Error occurred

POST `/api/discussions/:id/intervention`

Send a human message to the discussion

Request:

{
  "content": "Your message"
}

POST `/api/discussions/:id/stop`

Stop the discussion

Response:

{
  "status": "stopped"
}

Building for Production

# Build both frontend and backend
pnpm build

# Start production server
pnpm start

The frontend will be built to frontend/dist and the backend to backend/dist.

Troubleshooting

API Key Issues

Ensure your .env file is in the root directory
Verify your API keys are valid and have sufficient credits
Check that there are no extra spaces or quotes around the keys

Port Already in Use

If port 3001 or 3000 is already in use:

# Change the backend port in .env
PORT=3002

# Or kill the process using the port
lsof -ti:3001 | xargs kill -9

SSE Connection Issues

Make sure the backend is running before starting the frontend
Check browser console for connection errors
Ensure no firewall is blocking the connection

TypeScript Errors

# Clean and reinstall dependencies
rm -rf node_modules backend/node_modules frontend/node_modules
pnpm install

AWS Bedrock Issues

Ensure your AWS IAM user has the following permissions:
- bedrock:InvokeModel
- bedrock:InvokeModelWithResponseStream
Verify the Bedrock models are enabled in your AWS account (EU West 1 region)
Check that AWS credentials are properly set in your .env file
Ensure the AWS_REGION matches where your Bedrock models are available

Architecture

Frontend: React 18 + TypeScript + Vite + Tailwind CSS
Backend: Node.js + Express + TypeScript (ES Modules)
State Management: Zustand
Real-time Communication: Server-Sent Events (SSE)
Streaming: Token-by-token with 50ms batching for performance

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Fork the repository
Create your feature branch (git checkout -b feature/amazing-feature)
Commit your changes (git commit -m 'Add some amazing feature')
Push to the branch (git push origin feature/amazing-feature)
Open a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

Built with OpenAI API
Built with Anthropic API
Built with AWS Bedrock
UI components styled with Tailwind CSS

Support

For issues, questions, or suggestions, please open an issue.

Made with ❤️ by the Shaharia Lab team

Name		Name	Last commit message	Last commit date
Latest commit History 2 Commits
backend		backend
frontend		frontend
.dockerignore		.dockerignore
.env.example		.env.example
.gitignore		.gitignore
.mcp.json		.mcp.json
LICENSE		LICENSE
README.md		README.md
docker-compose.yml		docker-compose.yml
package.json		package.json
pnpm-lock.yaml		pnpm-lock.yaml
pnpm-workspace.yaml		pnpm-workspace.yaml

Uh oh!

License

shaharia-lab/multi-llm-discussion

Folders and files

Latest commit

History

Repository files navigation

Multi-LLM Discussion System

Features

Tech Stack

Frontend

Backend

Quick Start

Prerequisites

Option 1: Using Docker (Recommended)

Option 2: Local Development

Environment Variables

Supported Models

OpenAI

Anthropic

AWS Bedrock

How It Works

Usage Guide

Starting a Discussion

During a Discussion

Discussion Flow

Project Structure

API Endpoints

POST /api/discussions/start

GET /api/discussions/:id/stream

POST /api/discussions/:id/intervention

POST /api/discussions/:id/stop

Building for Production

Troubleshooting

API Key Issues

Port Already in Use

SSE Connection Issues

TypeScript Errors

AWS Bedrock Issues

Architecture

Contributing

License

Acknowledgments

Support

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Sponsor this project

Uh oh!

Packages 0

Uh oh!

Languages

POST `/api/discussions/start`

GET `/api/discussions/:id/stream`

POST `/api/discussions/:id/intervention`

POST `/api/discussions/:id/stop`

Packages