Chained Together

Demo

About

The problem it solves Current dating apps revolve around the same style of users advocating for themselves, but we believe getting the audience involved is best way to find a match. We know that people enjoy watching reality tv shows and love getting obsessed in relationships so we’ve tapped into the same instinct to allow people a front row seat, with actual influence over the outcomes. Chained Together is the first decentralized dating ecosystem where your community is your wingman. When users aren’t Tending to their own messages, they can spend time thoughtfully voting on matches for others, making the quest for love actually fun. To further reduce bias, we implemented an AI model that proposes initial pairings that the community will later vote on. The collective intuition of the "Matchmaker DAO" combines social consensus with blockchain-backed accountability. Essentially, We are turning the search for love into a transparent, community-driven experience. Use of AI tools and agents AI is used in profile intake and bio generation. The user’s onboarding answers are fed to the Qwen 2.5 7B model running on 0G Compute, which generates a 2-3 sentence bio summary. AI is also used for compatibility scoring (matchmaking). We send two users’ bios to the same Qwen 0G model, which returns a JSON score 0-100 and compatibility reasons. A background process continuously generates scored match pairs and feeds them into a queue. On the server side we use the 0g serving broker SDK for authenticated requests.

How it's made

Challenges we ran into One specific hurdle we ran into while building the project was integrating Snapshot, which powers the DAO governance layer. The backend initially would not connect correctly because Snapshot requires a properly configured ENS domain. We resolved this by registering and configuring an ENS domain and updating our Snapshot space settings, which allowed proposals and voting to function correctly. Another major challenge involved JSON parsing from LLM responses. The model frequently returned malformed JSON, often wrapped in markdown blocks or containing unbalanced braces, which caused failures in our backend pipeline. To overcome this, we implemented a robust extraction and recovery system that strips markdown formatting, repairs JSON structure, and automatically retries generation using a correction prompt when parsing fails. Finally, we could not use a traditional DAO as it would force users to send ETH in order to vote. So we adopted Snapshot’s gasless EIP-712 voting infrastructure so users can participate in DAO governance without paying transaction fees. These solutions allowed us to stabilize ChainedTogether, resulting in a resilient three-layer architecture combining AI inference on 0G Compute, DAO governance through on-chain and Snapshot voting, and Supabase as the application state machine, with fallback mechanisms ensuring the app continues working even when individual services fail.