Unagi Network
DePIN of autonomously verified security cameras that give agents eyes to make onchain actions
Demo
About
The problem it solves Currently, security footage is not a public resource (ie if a house gets robbed, the neighbor house has no obligation to provide camera footage). Secondly, the camera footage is not verifiable (ie in a court case, doctored video footage can falsely put people in jail. Thirdly, there's no easy way to agentically access cameras in a widespread manner. Unagi solves these issues by incentivizing people to provide their camera footage and make it accessible for agents to utilize and take action on. Unagi also uses Eigen Layer's AVS to verify these cameras before bringing them onchain, which solves problem #2. Finally, we use agentic and vector search to query these cameras using natural human prompting to answer complex questions onchain. User Interaction and Data Flow There are two types of users: 1) Camera providers - People can upload their camera footage information (Camera ID, title, stream url, etc) by clicking an initialize camera button. Once the user does this, they also have to connect their Metamask wallet so that they can receive the NFT of their camera footage. 2) End Users - People can query a simple prompt, which will then trigger a vector search for the most appropriate camera, as well as an agent that will commit any actions described in the prompt (ie find the details of...). Also end users can access cameras directly through our map. Finally, both of our users have the ability to click a button that causes the agent to execute a smart contract function (in our demo, we gave the user the ability to file an insurance claim). Additionally, they have the ability to dispute the IP if they believe someone else is getting incorrectly rewarded for their footage. The project architecture and development process The solution is a web app that allows users to provide their camera footage and generate IP on their cameras AND/OR query the existing camera network using agentic frameworks. Product Integrations We utilized Othentic + EigenLayer to autonomously verify that the cameras were not doctored footage. We utilized story protocol. Our tech stack included React.js for the frontend, Flask backend, supabase for storing camera streaming snapshots and inference. Key differentiators and uniqueness of the project This projects key features and unique aspects lie in the autonomous verification of the cameras using Othentic and Eigen Layer, the incentive mechanism by giving users IP to their cameras and then rewarding them story protocol royalties, and using CDP agentkit to conduct onchain actions like filing an insurance claim. Trade-offs and shortcuts while building For future implementation, we wanted to address privacy concerns. We would first token gate camera footage access to users that provide zk proofs of police claim or some other important and sensitive information. This ensures that cameras are only being queried by users that truly need them and don't have malicious intent. Another future implementation would be to sell this to governments or museums, which could use our system to add an extra layer of verification and security to their existing surveillance network. Finally, in the future we would want to use USB TEEs that would plug into each camera and allow the attestation to occur off chain, then bring the proven cameras on chain. Additional Features N/A