superpowers-memory

To persist agent memory and context across sessions in PostgreSQL, you can log events at session start, during execution, and at session end to store outcomes. This episodic storage preserves workflow context and enables audit trails for future retrieval.

Hydrating prior context when starting a new agent workflow involves retrieving relevant memories and insights from previous sessions at initialization. This surfaces key decisions and historical events to accelerate progress on ongoing tasks.

Yes, you can promote significant logged events into canonical insights and link them to source data. This process ties memories to their original events and tracks their status, creating a searchable knowledge base from workflow history.

PostgreSQL handles episodic memory storage, event logging, and cross-linking between events and insights with status tracking. Neo4j is referenced in the environment setup, but the core memory_add and hydration functions rely on PostgreSQL.

Standard logging often fails to preserve workflow context because it lacks structured episodic storage and hydration. Without logging outcomes at session end and retrieving prior memories at session start, agents lose historical decisions and audit trails.

PostgreSQL episodic storage for agent memory requires applying memory logging at session start, checkpoints, and end. If the workflow lacks explicit memory_add calls at these stages, context hydration will be incomplete and audit trails will have gaps.