RadarDrop AI Agent
Last updated
Last updated
Finding information about airdrop projects is not easy. Understanding crypto keywords, digging into documentation, and searching for posts on social networks can quickly become difficult and overwhelming.
β‘οΈ Introducing our AI agent, expertly trained on each airdrop with dozens of sources! Quickly access crucial information in any language and stay ahead of the game.
ποΈ Fast: We use the latest, most performant models from OpenAI and Mixtral to answer user requests.
π Accurate: By leveraging multiple data sources, our AI agent provides precise and up-to-date information about airdrop projects, ensuring users have all the details they need.
π Multilingual: The AI agent supports multiple languages, making it accessible to a global audience.
From the userβs text input to the modelβs response, there are a few steps:
We use WebSocket to ensure communication between the client and the server. Since the server instances are scalable, only one instance starts the WebSocket server.
Once the user has joined a channel (linked to an airdrop or not), he can send his message through the WebSocket connection. The message is then almost instantly sent to our queue system from BullMQ.
We can handle up to 100 messages simultaneously with our current infrastructure. We can easily increase this number to meet demand.
Once the relevant documents are gathered, we use our pre-trained AI models to understand the context and generate a coherent and accurate response. The response is then streamed back to the user through the WebSocket connection, ensuring real-time communication and quick resolution.
Once the request is handled by our chat service, we gather all the split documents from the vector database.
Those documents were previously fetched from social networks and the web. The documents were then split using our homemade splitters and embedded using .