In previous articles we already covered the vision, swarms and interfaces, but what does Hivemind actually do when you give it a prompt? In this deep dive we take a look at the data pipeline, and explain how the different components make this the most sophisticated Web3 discovery tool on the market.
Hivemind consists of multiple agent swarms, and each swarm covers one project, chain or ecosystem. A swarm consists of multiple agents, each with their own specialty. Let’s take a better look at the way each of these agents gathers data or information.
The agents inside the Hivemind
Through a certain interface, either built by DappRadar or third-party developers, users get to engage with Hivemind. Let’s assume that the interface allows users to input their prompts or questions. The prompt then first arrives at the Master agent, which serves as a central coordinator that manages all other agents. It maintains core knowledge, and delegates specific queries to the specialized agents. After obtaining all the information, it shares the response with the user. This can for example happen on X, in a chatroom, or through another type of interface.
As mentioned, the Master agent delegates specialized tasks to the other agents. Each of these agents has a different role in the swarm. Below we take a glimpse at some of the possible types of agents, and their capabilities.
Crawler agent
Processes and analyzes web content from project-related URLs. It then extracts relevant information to present to the user. At the same time, it maintains a cache of recently crawled pages to provide the Master Agent with up-to-date knowledge from the web. It can also extract transcripts from YouTube videos, allowing it to pull insights from AMAs, game guides, token analysis, and other video-based content.
Knowledge agent
This agent serves as a central knowledge repository for verified information about a project or ecosystem. It systematically processes and categorizes information about product features, mechanics and recent updates. While doing so, the Knowledge agent maintains strict verification standards to ensure information remains accurate.
X agent
The X agent monitors the project’s official account and community discussions on X. This agent also tracks important announcements from verified accounts, and analyzes social media trends. It’s even capable of gathering specific posts on demand. Ultimately the X agent needs to provide real-time updates to the Master agent about project information and the conversations that happen within the community.
Discord agent
As you may have guessed, the Discord agent gathers messages from the most important channels in a project’s official Discord server. For example, it may gather information from the #announcements, #updates and the #trading channel inside a certain Discord server. The Discord agent gathers the information and converts it into knowledge, which is then shared with the Master agent.
All these agents work together. The Master agent coordinates and sends tasks to the underlying agents in the swarm. These then provide the requested information, data and context, after which the Master agent analyzes it all to come up with a reply to the user’s request.
In the example painted above, the AI agent swarm consists of 4 agents and a Master agent. However, every AI agent swarm may consist of more or less agents, depending on the complexity of the project and the type of functions it would require.
Closing words
Hivemind’s pipeline transforms Web3’s chaos into clarity. From Crawler to X agents, each swarm meticulously gathers, verifies, and contextualizes data, delivering intelligent answers via the Master agent. This dynamic system adapts to any project’s complexity, ensuring precise, real-time insights. Whether through DappRadar or third-party interfaces, Hivemind empowers users with unparalleled Web3 discovery.
