Shipped memory-quality benchmark + eval harness, the TS counterpart of the
Python actrone_memory.benchmark, over the same LongMemEval-style dataset.
Runs offline (in-memory store + local embedder), so memory quality is a
measurable, CI-gated property in both libs (the "one shared eval" decision).
Honest by construction: default scores reflect the dependency-free
LocalEmbedder (keyword overlap). Pass a real embedder via runEval({manager})
to measure semantic recall.
Shipped memory-quality benchmark + eval harness, the TS counterpart of the Python
actrone_memory.benchmark, over the same LongMemEval-style dataset. Runs offline (in-memory store + local embedder), so memory quality is a measurable, CI-gated property in both libs (the "one shared eval" decision).Honest by construction: default scores reflect the dependency-free
LocalEmbedder(keyword overlap). Pass a real embedder viarunEval({manager})to measure semantic recall.