Skip to content

revsmoke/cerberus

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Cerberus: Multi-Model AI Collaboration System

A ColdFusion-based application that enables multiple AI models to work together as a cohesive team, providing richer, more nuanced responses than any single model could offer alone. Cerberus currently integrates:

  • Claude Opus 4 (Anthropic)
  • o4-mini (OpenAI)
  • Grok 3 Beta (X.AI)

Vision!

Like the mythological three-headed guardian, Cerberus represents a vigilant, multi-faceted intelligence system that can approach problems from different angles simultaneously. The ultimate goal is to create a unified "AI council" or "Greek chorus" that functions as a team of advisors with diverse perspectives and capabilities.

Rather than isolated systems that operate independently, Cerberus enables these models to:

  • Collaborate on complex problems by sharing information and insights
  • Complement each other's strengths and compensate for individual limitations
  • Cross-validate conclusions by providing multiple viewpoints
  • Combine specialized knowledge domains for more comprehensive answers

Features

  • Advanced Cross-Model Communication: Models can query each other for information, analysis, and perspectives
  • Visual Distinction: Color-coded messages differentiate between user messages, model responses, and cross-model communication
  • Anti-Recursion Protection: Sophisticated mechanisms prevent infinite loops in cross-model conversations
  • Conversation History: Per-model conversation tracking with unified backend storage
  • Side-by-Side Comparison: View all three models' responses simultaneously for easy comparison
  • File Handling: Upload and share images, PDFs, and other documents across models
  • Proxy Architecture: CORS-compliant ColdFusion proxy for API communication

Technical Architecture

  • Backend: ColdFusion 2023 with component-based API proxying
  • Frontend: Pure HTML, CSS, JavaScript (no external frameworks)
  • APIs: Anthropic, OpenAI, X.AI with unified response handling
  • Database: SQL storage for conversation history and context
  • File System: Local storage for uploaded files

Future Development

Cerberus is evolving toward a true consensus-building system with these planned enhancements:

  • Unified CFC Proxy: Refactoring to route all model communication through ColdFusion components
  • Consensus Framework: Structured yet flexible process for models to explore topics and converge on collective insights
  • Shared Knowledge Base: Session-persistent memory accessible by all models
  • Specialized Role Assignment: Dynamic allocation of roles to models based on query characteristics
  • Meta-Coordination: Orchestration layer for complex query handling across models

For detailed development roadmap, see CLAUDE.md in the repository.

Setup

  1. Install ColdFusion 2023
  2. Place files in your ColdFusion wwwroot directory
  3. Access the application at: http://localhost:8500/cerberus/index.cfm
  4. For enhanced capabilities, ensure database connectivity is configured

API Keys

The application requires API keys for each model:

  • Anthropic Claude: x-api-key header
  • OpenAI: Bearer token authentication
  • X.AI: Bearer token authentication

Store your API keys securely in the ClaudeCFC.cfc component.

Why Cerberus Matters

The development of Cerberus represents an important step forward in AI assistance technology. By enabling multiple models to collaborate rather than compete, we can:

  1. Reduce individual model biases through cross-validation and perspective sharing
  2. Leverage specialized capabilities of different model architectures
  3. Create emergent intelligence that exceeds what any single model could produce
  4. Provide more balanced, nuanced responses to complex queries
  5. Explore the potential of non-hierarchical AI collaboration as a new paradigm

Security Considerations

  • API keys are stored securely on the server side in the ClaudeCFC.cfc component
  • File uploads are validated for type, size, and security risks
  • Cross-model communication is monitored to prevent infinite loops
  • User input is properly sanitized before processing

License

MIT

Contributing

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

Key areas where contributions would be especially valuable:

  • Implementing the unified CFC proxy architecture
  • Enhancing the consensus-building framework
  • Improving the UI for better visualization of cross-model collaboration
  • Adding support for additional AI models
  • Creating specialized tools for specific knowledge domains

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors