[AI Integrations / AI Collab] - [Apr 12th] - [CineAI-Agents 🎬🤖] - Encode AI Hackathon

[manudev97]

[AI Integrations / AI Collab] - [Apr 12th] - [CineAI-Agents 🎬🤖] - Encode AI Hackathon

Repository Link

[Link to repository](https://github.com/manudev97/CineAI-Agents)

Contributors

[@manudev97](https://github.com/manudev97)

Accomplished Tasks

• Agent Creates a Space
• Agent Uploads onto Storacha
• Decentralized Knowledge Management

Update

• CineAI-Agents is an innovative project that combines artificial intelligence, decentralized storage, and audiovisual content analysis. Designed specifically to serve a Telegram community of over 4,000 movie enthusiasts, this project aims to provide faithful, unbiased summaries of films and TV series that members want to watch. By leveraging AI analysis of critic reviews, CineAI-Agents offers valuable insights without spoilers, helping community members make informed viewing decisions.
  • LangChain Agent (Web Scraper): Analyzes Filmaffinity reviews and stores data on Storacha
  • ElizaOS Agent (Telegram Bot): Interacts with users and retrieves knowledge from Storacha
• Decentralized Knowledge Base:
  • Successfully implemented FAISS index storage on Storacha
  • Review summaries are stored and accessible through decentralized infrastructure
  • Both agents can access and utilize the same knowledge base
• What went well:
  • Storacha integration for both text summaries and binary FAISS indexes
  • Cross-agent knowledge sharing through decentralized storage
  • Efficient vectorization and retrieval of movie review information
• Challenges faced:
  • Implementing binary file uploads to Storacha
  • Errors when uploading files with the Elisa plugin to Storacha
  • Synchronizing knowledge between the two agents
  • Optimizing vector search for real-time applications

Demo Videos

Feedback

• Architectures and UCAN Integration: Due to time constraints, we implemented a basic architecture where our application owns the Space. In future iterations, we'd like to explore more sophisticated options:
• Client-server: Currently, our application owns the Space, users interact with our agents, and we handle uploads to Storacha.
• Delegated: We would like to explore providing our users with delegations that allow them to upload their own movie reviews directly to our Space.
• User-owned: For film critics or professional reviewers in our community, we're interested in enabling them to own their own Space and upload directly, with our agents consuming and organizing their content.

Having clearer documentation on implementing UCAN delegations would be extremely helpful for moving to these more advanced architectures.

Next Tasks

• Implement a more robust synchronization mechanism between agents
• Add more data sources beyond Filmaffinity (IMDB, Rotten Tomatoes, etc.)
• Create a recommendation system based on stored reviews
• Develop a user preference learning module for the Telegram bot
• Implement multi-language support for both extraction and interaction
• Build a web interface for directly querying the review database
• Add support for TV shows and series analysis with episode-level granularity
• Optimize FAISS index for faster retrieval on larger datasets
• Explore delegated and user-owned UCAN architectures for more direct community contribution