Add Cache flag to api (#1642)

api/types.go

@@ -103,6 +103,7 @@ 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
@@ -355,6 +356,7 @@ func DefaultOptions() Options {
 		MirostatEta:      0.1,
 		PenalizeNewline:  true,
 		Seed:             -1,
+		Cache:            true,
 
 		Runner: Runner{
 			// options set when the model is loaded

docs/api.md

@@ -306,7 +306,8 @@ curl http://localhost:11434/api/generate -d '{
     "embedding_only": false,
     "rope_frequency_base": 1.1,
     "rope_frequency_scale": 0.8,
-    "num_thread": 8
+    "num_thread": 8,
+    "cache": true
   }
 }'
 ```

llm/ext_server.go

@@ -235,6 +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,
 	}
 
 	if predict.Format == "json" {