Frames Autonomous Recommendation Management Service (FARMS). Leverage AI to create and deploy frames, and also Personalize your farcaster(warpcast) feeds
Prize Pool
FARMS - (Frames Autonomous Recommendation Management Service) is a tool to automate Frames Creation and Personalize content feeds using AI Assistance in a simple Web App. We allow user to upload and create a frame with open framework via files they own but also to prompt and create a fully functioning frame from a single sentence using LLMs and Agent Technology. They can also view their neighbours and followings casts and using a simple prompt, an AI will order and personalize the frames feed to exactly how they describe.
Using Dynamic to login via web2/web3 and reduce friction. We also offer Privy Login also
When a user logs in they see our dashboard which is made with react and designed using tailwind. We then split the screen and offer the 2 features of FARMS.
First to automate frames creation and deployment using AI (unfinished) We had designed a sandbox design space on the front end which allows drag-n-drop of images and custom components to design your frames. You can then use the ai prompt to ask it to complete and use your uploaded components, or ask it to create a totally new frame from the beginning. This is all routed to our backend server, and run through our python scripts running OpenAI chatGPT3/4 and also pydantic for formulating the correct component formats in order to translate from LLM output -> Structured frames (Open Framework) format. Also frames.js is used to construct all frames here.
The second feature is Personalization feeds. What we do is first load the users followings and closest neighbours on farcaster using Karma3Labs API and then Pinata API for casts/frames information. We then render this for our users in a generic way. The user can then prompt using our chat box, an AI to personalize and filter/order their generic feed in any way they want. We route this through OpenAI ChatGPT again and use Pydantic for valid input and output formats in order to sort and filter the frames. This is returned back to the front end and we also recognize the need for interoperability for this personalized feature to other apps, Dapps and tools, so we expose an api url for that users specific personalized feed, which the user can copy and share, so others can retrieve their personalized Frames feed.