How does AURA work?
AURA operates through a streamlined and efficient pipeline: Account Address → AdEx Velcro → LLM → Security Processing → Inverse Humanizer. Here’s how it works in detail:
Account Address: The process kicks off when the user inputs their account address.
AdEx Velcro: Developed in collaboration with Ambire Wallet, Velcro optimizes data extraction by querying the network’s RPC. This allows AURA to quickly gather a variety of portfolio data points—such as ERC20 tokens, NFTs, and their prices—in a single call (eth_call). This enables fast access to token balances, asset prices, and NFT holdings.
LLM (Large Language Model): AURA then converts the portfolio data into a unique prompt, which is sent to the LLM for processing. The model uses this information to generate personalized insights and recommendations, helping users spot relevant opportunities. AURA continuously updates itself by integrating a variety of data sources, including Web3 app aggregators like DeFiLlama and DappRadar, airdrop hunting APIs, and onchain analysis tools, ensuring that its suggestions are always up-to-date. In its initial version, AURA focuses on analyzing users' portfolios, but future updates will expand to offer deeper insights into user behavior, such as identifying trading patterns or yield farming strategies. AURA will also factor in previously held assets, like membership NFTs, to further refine its understanding of user preferences.
Security Processing: After the LLM generates insights, the data undergoes a security screening to protect against potential risks, such as LLM poisoning attacks. This step ensures that the user's data remains secure throughout the process.
Inverse Humanizer: Finally, the processed data is converted from human-readable insights back into machine-readable instructions, such as transactions or intents. This allows the system to take the necessary actions on the user's behalf through an integrated UI or AI agent.
Last updated