A community-band library app is a short Python script; the ticketing service is another backend that already speaks gRPC. Both need the same Triton-served encoder. How should each client connect?
Select an answer to reveal the explanation.
Short Explanation
Same Triton encoder, two clients. The short Python script can speak REST or HTTP. The ticketing backend already speaks gRPC. The protocol does not change model quality, and a per-client CUDA kernel is not how you pick a transport.
Full Explanation
Triton exposes both HTTP and gRPC for the same served model, so the simple Python script can use REST or HTTP while the ticketing backend keeps gRPC. HTTP is a supported Triton client path, not a blocker. The transport does not change encoder quality. A per-client CUDA kernel is not how associate apps choose a protocol.