Does batch processing support it? (like OpenAI, use batch to reduce cost when call embedding model) <img width="947" height="323" alt="Image" src="https://github.com/user-attachments/assets/901a6eeb-8fa5-4d72-9716-8fff335a9178" />