memory-search

You search agent memory for past conversation history by applying QMD semantic search to query stored knowledge, utilizing BM25 keyword matching, vector similarity, or a combined query with reranking for accurate recall.

Semantic search finds information in a knowledge base by analyzing the meaning and context of your query using vector similarity, allowing you to discover relevant information even when exact keywords do not match.

You can filter memory searches to specific collections like workspace knowledge or past conversations, narrowing down the scope of your query to target specific areas of stored information.

The best way to recall learned information from agent session history is using a combined query with reranking, which fuses BM25 keyword search and vector similarity to return the most accurate and relevant results.

Keyword search works for recalling past conversations if you remember specific words, utilizing traditional BM25 keyword matching to quickly find information based on exact term matching within the stored memory.