Show HN: I built an AI agent memory engine because Obsidian wasn't cutting it (perseus.observer)

Thomas Connally of Perseus Computing has developed Perseus Vault, an AI agent memory engine designed to address the limitations faced by agents using traditional note-taking tools like Obsidian. After encountering significant challenges in retrieval and context management when scaling from 2,000 to over 10,000 notes, Connally identified that AI requires a structured memory system distinctly different from human-centric applications. Perseus Vault is a compact, offline solution that emphasizes efficient search, prioritization of information, deduplication, and historical accuracy of facts—all crucial for enhancing AI agent performance. Significantly, Perseus Vault employs advanced features like hybrid search combining keyword and vector approaches, a weighted memory decay system, and bi-temporal history to ensure relevance and up-to-date information for AI agents. It is designed as a single Rust binary, compatible with various agents and operates independently of cloud services, making it particularly appealing for industries such as defense and healthcare where data privacy is critical. With this innovation, Connally aims to transform how AI agents manage contextual knowledge, enhancing their reliability and the efficiency of their interactions.