pub fn embed_entity_texts_cached(
models_dir: &Path,
texts: &[String],
parallelism: usize,
) -> Result<(Vec<Vec<f32>>, EmbedCacheStats), AppError>Expand description
G56: embeds entity-name texts through a process-wide cache.
Skips any (model, text) pair already produced in this CLI invocation
and only spawns subprocesses for the cache misses. Returns vectors in
the same order as texts.
Designed for entity-name batches (short texts). For chunk embeds use
embed_passages_parallel_local directly — chunks are unique per
memory and cache hit rate is negligible.