api: add generate endpoint for structured outputs (#7939)

api/types.go

@@ -67,7 +67,7 @@ type GenerateRequest struct {
 	Raw bool `json:"raw,omitempty"`
 
 	// Format specifies the format to return a response in.
-	Format string `json:"format"`
+	Format json.RawMessage `json:"format,omitempty"`
 
 	// KeepAlive controls how long the model will stay loaded in memory following
 	// this request.

cmd/cmd.go

@@ -1126,7 +1126,7 @@ func generate(cmd *cobra.Command, opts runOptions) error {
 		Prompt:    opts.Prompt,
 		Context:   generateContext,
 		Images:    opts.Images,
-		Format:    opts.Format,
+		Format:    json.RawMessage(opts.Format),
 		System:    opts.System,
 		Options:   opts.Options,
 		KeepAlive: opts.KeepAlive,

server/routes.go

@@ -148,10 +148,7 @@ func (s *Server) GenerateHandler(c *gin.Context) {
 		return
 	}
 
-	if req.Format != "" && req.Format != "json" {
-		c.AbortWithStatusJSON(http.StatusBadRequest, gin.H{"error": "format must be empty or \"json\""})
-		return
-	} else if req.Raw && (req.Template != "" || req.System != "" || len(req.Context) > 0) {
+	if req.Raw && (req.Template != "" || req.System != "" || len(req.Context) > 0) {
 		c.AbortWithStatusJSON(http.StatusBadRequest, gin.H{"error": "raw mode does not support template, system, or context"})
 		return
 	}
@@ -278,7 +275,7 @@ func (s *Server) GenerateHandler(c *gin.Context) {
 		if err := r.Completion(c.Request.Context(), llm.CompletionRequest{
 			Prompt:  prompt,
 			Images:  images,
-			Format:  json.RawMessage(req.Format),
+			Format:  req.Format,
 			Options: opts,
 		}, func(cr llm.CompletionResponse) {
 			res := api.GenerateResponse{