Show HN: Oc-mnemoria – Persistent memory for AI coding agents (github.com)

A new project called oc-mnemoria has been announced, designed to provide persistent memory for AI coding agents within the OpenCode environment. By introducing a shared "hive mind," it allows agents to retain context across sessions, enabling them to read and write to a single memory store. This approach ensures that agents can access relevant information compiled by other agents (e.g., plans, decisions, bug fixes) without losing continuity, which is crucial for enhancing collaboration and efficiency in AI-assisted development. Significantly, oc-mnemoria employs a hybrid search engine, mnemoria, built in Rust, which combines BM25 and semantic search capabilities while ensuring data integrity through CRC32 checksums and corruption recovery. The memory is stored locally in an append-only binary format, fostering a version-controlled environment that can be audited alongside code changes. This persistent memory solution promises to improve the productivity of AI agents, allowing them to build on previous insights and decisions, thus streamlining coding processes and reducing redundancy in tasks.