Memory Schema

Handles schema lifecycle for Basic Memory's structured notes. You run schema_infer on a note type like Meeting or Person to generate a schema from existing notes, write it to the schema directory using Picoschema syntax (think YAML with types, optionals, enums, and relations), then use schema_validate to check conformance. The validation looks for fields as observation categories in note bodies, not just frontmatter, which is a quirk worth knowing. Run schema_diff periodically to catch when your notes evolve faster than your schemas. Works in warn mode by default so it won't block writes, just complains. Good for keeping a knowledge graph consistent without fighting it.