memory

Store persistent memories for projects across sessions in Supabase by writing entries to a centralized memories table using a defined schema. This enables cross-session continuity by applying controlled write, read, and search operations for project, decision, and knowledge data.

Managing multi-user memory persistence requires per-user scoping and key-based uniqueness within a centralized database. This approach enforces data storage rules and governance, ensuring explicit provenance and preventing local file writes across different projects.

Deprecate obsolete memories instead of deleting them by applying self-modification rules that merge duplicates and assign high, medium, or low confidence based on explicit state or inferred signals. This lifecycle management prevents data loss.

Yes, you can derive a project slug automatically from a git repository. The system derives the project slug from the git origin or current directory and normalizes it to a lowercase hyphenated form for consistent memory categorization across sessions.

Yes, this memory management approach supports text search across stored values. You can search by text across memory values with recency consideration, and read or query data specifically by category and subcategory to retrieve relevant project decisions.

Writing persistent memory data to local files is prohibited by the data storage governance rules. Memories must never be written to local files; instead, they are created, updated, and searched through controlled operations within a Supabase memories table for centralized durability.