Hot AI Integrations, Hotter Storage & Spicy AI Collaboration with Storacha Data Storage - April 13, 2025 - NexaMind - Encode AI Hackathon

[KPR-V]

Hot AI Integrations, Hotter Storage & Spicy AI Collaboration with Storacha Data Storage - April 13, 2025 - NexaMind - Encode AI Hackathon

Repository Link

https://github.com/KPR-V/nexamind

Contributors

@KPR-V
@Sehaj-0186

Accomplished Tasks

• Implemented a fully functional AI chat platform with multiple LLM models
• Created AI image generation capabilities with Lilypad API integration
• Built decentralized storage system using Storacha for conversations, images, and files
• Developed wallet-based encryption for secure data storage
• Added web search capabilities for up-to-date information in AI responses
• Designed responsive UI for both desktop and mobile devices
• Created a dashboard to view and manage stored content
• Implemented AI Agents System with specialized roles and capabilities
• Added multi-agent collaboration for solving complex tasks
• Integrated document-based context for more informed AI responses
• Built LlamaIndex integration for advanced retrieval-augmented generation (RAG)
• Developed custom tools system for extending agent capabilities

Update

• NexaMind is an all-in-one AI assistant platform that combines multiple AI capabilities with decentralized storage
• The project integrates Lilypad API to provide access to various LLM models and image generation capabilities
  • Successfully implemented multiple model selection for both text and image generation
  • The web search tool enhances AI responses with up-to-date information
• AI Agents System is now a core feature of the platform
  • Users can create and customize specialized AI agents with specific roles and expertise
  • Agents can be equipped with various tools including web search, document analysis, and more
  • The multi-agent collaboration system allows different agents to work together on complex tasks
  • Document upload capability enables context-aware responses based on user-provided materials
• LlamaIndex integration provides powerful RAG capabilities
  • Automatically indexes and embeds uploaded documents for contextual retrieval
  • Supports both OpenAI and local Ollama models for flexibility
  • Custom query engines for more precise information retrieval
• Storacha integration remains a core feature of the platform
  • When a user connects their wallet, their wallet address is used to create a dedicated space in Storacha
  • All conversations, generated images, uploaded files, and agent configurations are automatically encrypted and stored
  • The encryption is based on the user's wallet address, ensuring only they can access their data
  • The implementation provides a seamless decentralized storage experience without compromising security
• Technical challenges included:
  • Implementing proper encryption/decryption for different data types
  • Managing asynchronous storage operations while maintaining UI responsiveness
  • Ensuring wallet-based authentication worked consistently across sessions
  • Orchestrating multi-agent workflows and handling complex document contexts
  • Managing state across collaborative agent tasks
• User experience was prioritized throughout development
  • Intuitive agent creation and management interface
  • Auto-saving of conversations and generated images reduces friction
  • The dashboard provides a clean interface to access all stored content
  • Agent workspace designed for ease of use and powerful functionality

Demo Videos

(please watch this at 0.5x for better understanding)

Gitbook

Nexamind Detailed docs

Architecture Highlights

• Agent System: Built on LlamaIndex with custom agent orchestration
• Knowledge Base: Document processing using vector embeddings for context retrieval
• Tool Framework: Extensible system for adding new capabilities to agents
• Decentralized Storage: Integration with Storacha for permanent, encrypted data storage
• Model Flexibility: Support for both cloud-based (OpenAI) and local (Ollama) LLM models

Feedback

• The lilypad integration was a bit hard for the streaming responses
• The Storacha integration process was a easy task
• Documentation on how to retrieve different spaces created in Storacha would improve the developer experience
• We'd appreciate any feedback on our implementation of wallet-based spaces for user data
• The LlamaIndex integration provides powerful capabilities but requires optimizations for edge devices

Next Tasks

• Implement collaborative spaces where multiple wallet addresses can access shared content
• Create a plugin system to extend AI capabilities with custom tools
• Implement token-gated access to premium AI models
• Add support for agent memory persistence across sessions
• Enhance the query optimization for better RAG performance
• Implement streaming responses for collaborative agent workflows
• We are interested in extending this project and would like to participate in the grants program