Octoplorer - block query made easy

For many Web3 users, we are certain that many people have heard the phrase "the data is on the blockchain, so everybody can see it." But how many have actually tried to do it, and then realized how difficult it is to query the data on the blockchain when it is more than just 1 transaction? We have created the solution for that problem: We make querying data on the blockchain (for now Ethereum and Polygon mainnets) easy using AI-powered natural language understanding and a friendly user interface for Web3 technology.

Members of our teams have various potential use cases for this web app:

• Obtain NFT images owned by a certain entity. This could be a
• In terms of number of follower, get the Lens handles for the top 5 most popular ids.
• Obtain the aggregate statistics on certain NFT collections on popular marketplaces (for now Opensea and Rarible)
• Obtain data for recent transactions within the past 7 days for a certain wallet
• For an Ethereum Name Service name (for example, "dwr.eth"), we can get the Farcaster name, Farcaster account details, connected address, and all token balances and images

As of project submission deadline, please keep in mind that we cannot address all of the above-mentioned use cases due to time limitation. With our best effort, we have been able to make querying the blockchain both an easy-to-understand and enjoyable experience through some simple use cases. User can also send the query's result to any recipient with a Polygon address through simple push messaging.

Basically, the more protocols and data that we plug in, the more possibilities will open up for us to query.

We also think that this simple yet user-friendly web app would help to onboard the next million new people into web3 while empowering data geeks with the information they need for their web3 journey.

*Side notes and trivia: for those that may be wondering where the name Octoplorer came from, it is a combination of the words "octopus" and "explorer". I (Lam) chose octopus as the representative animal because:

• 🐙 Octopus is fast, it can swim up to 25 miles per hour. That is faster than Michael Phelps!
• An octopus can multitask well due to its large nerve cluster. It has something like a "minibrain" at the base of each arm to independently control its movement. That's why it can grab multiple blockchain transactions simultaneously!

For the ideation stage, Lam literally drew the wireframe of the website on paper. Then Andrey used Figma to draft up a mock layout of the website along with the Octoplorer logo. For the blockchain data, we are thankful for the support of the sponsors Airstack and Lens Protocol. Thanks to Airstack's powerful up-to-the-minute API, we could query blockchain data on both Ethereum and Polygon mainnet. For Lens Protocol (on Polygon), we can also query many Lens-related identities such as handles, id, number of followers. For the natural-language query, we use OpenAI's ChatGPT 3.5 "text-davinci-003" model, enhanced by the langchain (open-source) library to empower the AI even further. When a user writes a query on the front-end, we have an large-language-model (LLM) agent that would decide whether that is a Len-related query or an Airstack-related query. After the LLM agent decides the model, then the query would be routed to the right ChatGPT model to interpret the query and write the appropriate GraphQL code. From there, the GraphQL code would then be passed to the right API endpoint (Airstack or Lens) to obtain the data necessary. After the data is obtained, we will then display the results back to the user in an easy-to-understand format, including tables, images, and/or graphs. For the front-end, we used Next.js Tailwind and React. We have also integrated web3 wallet sign-in (including Metamask) using "wagmi" and rainbow kit libraries. Thanks to the sponsor Push, we have also enabled a messaging service that allows users to send the results of their queries to any recipient on Ethereum mainnet without leaving the application. Since any LLM is non-deterministic by the nature of the model and GraphQL code for a specific protocol is strictly conformant to that protocol's unique schema, we did some hacky and creative things in order for some of the queries to work consistently. Please feel free to ask us about this if you are curious. :)