enable cache_prompt by default

api/types.go

@@ -103,7 +103,6 @@ type Options struct {
 	MirostatEta      float32  `json:"mirostat_eta,omitempty"`
 	PenalizeNewline  bool     `json:"penalize_newline,omitempty"`
 	Stop             []string `json:"stop,omitempty"`
-	Cache            bool     `json:"cache,omitempty"`
 }
 
 // Runner options which must be set when the model is loaded into memory

llm/ext_server.go

@@ -235,7 +235,7 @@ func predict(llm extServer, opts api.Options, ctx context.Context, predict Predi
 		"seed":              opts.Seed,
 		"stop":              opts.Stop,
 		"image_data":        imageData,
-		"cache_prompt":      opts.Cache,
+		"cache_prompt":      true,
 	}
 
 	if predict.Format == "json" {