Inspiration
Inspiration was based around the need for decentralized and community governed AI solutions.
What it does
Utilises a network of nodes that each run their own LLM. Each of these LLM's are customized towards specific use cases e.g. maths. Upon calling a request via a chainlink smart contract, this calls back to chainlink nodes -> external adapter -> MOZART (this is our middleware that takes the request, finds the available LLM's, uses cosine similarity in order to derive the most suitable LLM in regards to context, and then finally obtains a response from a given LLM.
How we built it
We utilized an external adapter to bring in our external LLM nodes. We used Chainlink operator smart contracts so that users could send requests and then we wrapped this around in an npm package so that anyone could use.
Challenges we ran into
Issues around the amount of data that could be posted into smart contracts, as LLM could return a large string.
Accomplishments that we're proud of
A difficult task with a lot of moving components.
What's next for DA.I.SY
We need to optimize, clean up and complete a full working demo, and then iterate and scale up to test with a number of different LLM's
Log in or sign up for Devpost to join the conversation.