Agents-MCP-Hackathon/SocialNetwork

Possibly the World’s First Social Network with LLMs as the Interface

🎥 Demo Video (Claude Desktop Client)

🛠️ Tech Stack

• —Modal
• —multilingual-e5-large (for text embeddings)
• —PostgreSQL + pgvector
• —Gradio MCP

🤔 Why This Project?

Most modern social networks encourage passive consumption through endless feeds, leading to mindless scrolling and reduced focus. This Social Network flips that model.

Here, a Large Language Model (LLM) acts as an interface to the network. Users interact through purposeful search and conversation, meaning:

• —You must articulate your intent.
• —You’re prompted to think critically and consciously about what you want.
• —The LLM can help guide, suggest, or even expand your thinking — making the experience more mindful, creative, and social.

It’s not just a platform. It’s a shift in how we connect.

🧠 Benefits

• —Psychologically healthier: No infinite scroll, no algorithmic noise.
• —Intention-based discovery: You only see what you seek.
• —Enhanced creativity: The LLM becomes a thought partner — not a distraction.
• —Noise-free socializing: No unsolicited news, drama, or content.

🚀 Features

• —Intelligent Posting: Users can publish text posts, which are embedded into a 1024-dimensional vector space using the multilingual-e5-large model.

• —Semantic Search: Posts are retrieved based on semantic similarity. Query text is embedded and compared using L2 distance to return the most relevant posts.

• —Comments: Users can engage with posts through comments.