Fix prompt caching on llama.cpp endpoints (#920)

Explicitly enable prompt caching on llama.cpp endpoints

Co-authored-by: Nathan Sarrazin <[email protected]>

src/lib/server/endpoints/llamacpp/endpointLlamacpp.ts

@@ -41,6 +41,7 @@ export function endpointLlamacpp(
 				stop: model.parameters.stop,
 				repeat_penalty: model.parameters.repetition_penalty,
 				n_predict: model.parameters.max_new_tokens,
+				cache_prompt: true,
 			}),
 		});