FirstBatch is a decentralized digital identity and personalization protocol that leverages AI and vector database technology to deliver real-time content recommendations while preserving user privacy. Its core mechanism uses user behavioral signals (e.g., likes, clicks) to dynamically update user embedding vectors without storing raw data, thus avoiding privacy leakage. The project provides an SDK supporting major vector databases like Pinecone and Weaviate, enabling developers to quickly integrate personalized recommendation features. The SDK includes modules for session management, signal collection, batch recommendation retrieval, and offers built-in recommendation algorithms with support for custom algorithms. The protocol aims to provide personalized experiences without centralized data collection for DApps, social platforms, the metaverse, and other scenarios.