High-level technical diagram
Last updated
Last updated
While Ball-E’s initial release emphasizes DeFi intelligence, its design is generalizable and can handle arbitrary tasks. Here is a high-level technical breakdown:
Intract Persona
User Data Consolidation: Ball-E obtains a holistic view by unifying on-chain, social, and preference data.
Behavioral Learning: Observes what tokens you frequently trade, which communities you engage with, and your typical risk appetite.
Short-Term and Long-Term Memory
Short-Term: Stores session context (recent queries or tasks). Ensures conversation continuity and quick references.
Long-Term: Maintains a historical record of your interactions, performance outcomes, and evolving preferences—securely stored and utilized in TEEs or privacy-preserving vaults.
Ball-E does not operate alone. It can hand off specific tasks to specialized on-chain agents (e.g., an advanced risk engine like “Eliza”) through:
Cross-Agent Communication: Standardized protocols (potentially leveraging smart contracts or direct agent-to-agent calls).
Collaborative Intelligence: Combining insights to maximize user benefits. For instance, Ball-E’s Alpha Agent might detect a new yield farm, then request the Risk Analyzer to evaluate it, and finally pass instructions to the Portfolio Manager to deploy capital.
TEE-Based Confidential Compute
Ephemerally decrypts your vault data for indexing or training, ensuring your private info remains protected.
Decentralized Identity & OAuth
Powered by Intract’s identity protocol, gating access to Ball-E’s functionalities based on user credentials (e.g., Rewind NFT, whitelisted addresses).
Tool & Action Interface
Each agent (e.g., Portfolio Manager) has a set of actions (move tokens, swap assets, post tweets) requiring user signature or on-chain transaction confirmations.
Ball-E requests user permission before executing or delegating any major action.