ReadonlydimensionsThe dimensionality of vectors this embedder produces.
Optional ReadonlyrelevanceThe cosine similarity at which this model's results turn from unrelated to relevant, used as the admission threshold when MemoryConfig.relevanceThreshold is not set. Declare it for a custom embedder once you have measured it; leave it out to use the library default.
Embedding seam.
MemoryManager.create()picks the best local embedder available (buildLocalEmbedder): FastEmbedEmbedder (bge-small-en-v1.5, semantic) when the optionalfastembedpackage is installed, otherwise the dependency-free LocalEmbedder, which hashes words into a bag-of-words vector so cosine similarity reflects shared keywords only. The lexical embedder is deterministic and offline, which also makes the test suite run anywhere.To use a hosted model, pass your own Embedder (for example one backed by OpenAI
text-embedding-3-small) and setrelevanceThresholdfor it.