Article URL: https://www.adaptiverecall.com/ Comments URL: https://news.ycombinator.com/item?id=48884815 Points: 15 # Comments: 0

Store, recall, and forget with a memory system that learns from every interaction. Retrieval quality improves automatically over time, powered by cognitive science and machine learning. Most memory APIs store embeddings and search by cosine similarity. Adaptive Recall does that and five layers more. Six capabilities that no other memory API offers, working together in every query. Four search strategies run in parallel: vector similarity, temporal recency, full-text keyword, and knowledge graph traversal. The system learns which strategies to prioritize for each type of query. Results are ranked using ACT-R activation modeling from cognitive science. Recency, access frequency, entity connections, and validated confidence all factor into which memories surface first. Entities and relationships are extracted automatically from stored memories. The graph becomes a retrieval pathway, finding relevant information through connections rather than just text similarity. Memories are not static rows in a database. They progress through stages, gain or lose confidence based on corroborating evidence, and fade naturally when no longer accessed. The system trains ML models on your usage data, validates every parameter change against real query history, and monitors its own retrieval quality. It gets better the more you use it.