RightMemory

Based on the available evidence, RightMemory looks promising, but it is better judged as an early open-source memory layer with a clear thesis rather than a broadly validated production-grade standard. The heat proof is limited: a small amount of GitHub stars/forks and a cluster of Zhihu posts show attention, not proven reliability. The stronger usefulness proof comes from the official repo description and longer author-written articles, which explain the architecture and intended workflow fairly clearly. However, there is still little third-party hands-on testing, benchmarking, or broad deployment evidence, so adoption should be cautious.\n\nIn practice, this is not a generic vector database and not just a private notebook for one chat assistant. A better analogy is a Git-friendly project knowledge layer for AI coding agents, with Markdown-backed tree + graph memory. Across the evidence, the intended outputs are persistent project decisions, user preferences, coding constraints, review learnings, command caveats, and other context that should survive sessions, clients, and devices.