polign-recall adds corrective typed memory for AI agents

Recall is a typed, durable memory and state library for AI agents that stores corrections and task state as append-only records in cloud object storage or databases. The system handles conflicting updates by marking previous values as replaced on read rather than modifying or deleting existing records in place. In comparative tests against mem0 and Zep over a nine-month scenario, Recall successfully answered all scenarios and retained dietary restrictions when asked to forget conflicting allergies. Open article →