Fix embeddings load model behavior (#2848)

api/types.go

@@ -121,7 +121,6 @@ type Runner struct {
 	VocabOnly          bool    `json:"vocab_only,omitempty"`
 	UseMMap            bool    `json:"use_mmap,omitempty"`
 	UseMLock           bool    `json:"use_mlock,omitempty"`
-	EmbeddingOnly      bool    `json:"embedding_only,omitempty"`
 	RopeFrequencyBase  float32 `json:"rope_frequency_base,omitempty"`
 	RopeFrequencyScale float32 `json:"rope_frequency_scale,omitempty"`
 	NumThread          int     `json:"num_thread,omitempty"`
@@ -395,7 +394,6 @@ func DefaultOptions() Options {
 			UseMLock:           false,
 			UseMMap:            true,
 			UseNUMA:            false,
-			EmbeddingOnly:      true,
 		},
 	}
 }

docs/api.md

@@ -321,7 +321,6 @@ curl http://localhost:11434/api/generate -d '{
     "vocab_only": false,
     "use_mmap": true,
     "use_mlock": false,
-    "embedding_only": false,
     "rope_frequency_base": 1.1,
     "rope_frequency_scale": 0.8,
     "num_thread": 8

server/routes.go

@@ -385,7 +385,7 @@ func GenerateHandler(c *gin.Context) {
 	streamResponse(c, ch)
 }
 
-func EmbeddingHandler(c *gin.Context) {
+func EmbeddingsHandler(c *gin.Context) {
 	loaded.mu.Lock()
 	defer loaded.mu.Unlock()
 
@@ -438,8 +438,9 @@ func EmbeddingHandler(c *gin.Context) {
 		return
 	}
 
-	if !loaded.Options.EmbeddingOnly {
-		c.JSON(http.StatusBadRequest, gin.H{"error": "embedding option must be set to true"})
+	// an empty request loads the model
+	if req.Prompt == "" {
+		c.JSON(http.StatusOK, api.EmbeddingResponse{Embedding: []float64{}})
 		return
 	}
 
@@ -948,7 +949,7 @@ func (s *Server) GenerateRoutes() http.Handler {
 	r.POST("/api/pull", PullModelHandler)
 	r.POST("/api/generate", GenerateHandler)
 	r.POST("/api/chat", ChatHandler)
-	r.POST("/api/embeddings", EmbeddingHandler)
+	r.POST("/api/embeddings", EmbeddingsHandler)
 	r.POST("/api/create", CreateModelHandler)
 	r.POST("/api/push", PushModelHandler)
 	r.POST("/api/copy", CopyModelHandler)