Size the reranker's TEI thread pools to its 1-CPU limit

Shipped

September 25, 2026 at 5:34 PM UTC

Author

Kamo

Commit

d25462f

The reranker runs embedding-model's TEI 1.9.4 under the same 1-CPU limit, so it takes the same sizing (KamoAI SP00 final review I-1): one tokenization worker, and OMP/MKL/RAYON at one thread. TEI's own defaults count host cores, 48 on k3m1, all inside a 1-CPU CFS quota. The image comment no longer says the embeddings instance is on 1.5; it moved to 1.9.4 in 5edb887.