Lens Alfred
Query all the LENS data in natural language! Discover content, find the accounts to follow, understand your followers!
Lens Alfred
Created At
ETHGlobal Lisbon
Winner of
🏊 Polygon — Build on Polygon
🏊 Scroll — Deploy a Smart Contract
Project Description
The LENS ecosystem contains a lot of social media and post data that is extremely difficult to discover. It feels like the post relevance function is whatever Stani mirrors. We realised we have the tools at our disposal to fix this so we got to work. We've created the ultimate natural language interface to the LENS ecosystem. Using Alfred, anyone can explore and discover interesting content.
See our demo video here: https://www.loom.com/share/ec262dfe0755449fbfe89f8d38a53201
How it's Made
A lot had to come together to make our vision a reality (in just under one day!):
- researching the initial use cases by interviewing our ideal users and then exploring what was feasible to build within 24hrs
- filtering and downloading the LENS data.
- generating embeddings for each of the posts and storing them in a vector db (Chroma). Not having access to a GPU, fully embedding all LENS posts would take many days on a macbook.
- iterating many times over the backend execution architecture, command format and the prompts to the LLM
- using and understanding external APIs including the GraphQL endpoints for LENS and Airstack
