server: add tests and fix isHuggingFaceURL edge case

- Add comprehensive tests for isHuggingFaceURL and getNumDownloadParts
- Fix bug where domains ending in huggingface.co (like nothuggingface.co)
  would incorrectly match as HuggingFace URLs
- Improve code comments with more detailed documentation

.github/workflows/release.yaml

@@ -372,13 +372,17 @@ jobs:
           outputs: type=local,dest=dist/${{ matrix.os }}-${{ matrix.arch }}
           cache-from: type=registry,ref=${{ vars.DOCKER_REPO }}:latest
           cache-to: type=inline
+      - name: Deduplicate CUDA libraries
+        run: |
+          ./scripts/deduplicate_cuda_libs.sh dist/${{ matrix.os }}-${{ matrix.arch }}
       - run: |
           for COMPONENT in bin/* lib/ollama/*; do
             case "$COMPONENT" in
-              bin/ollama)                echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
+              bin/ollama*)               echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
               lib/ollama/*.so*)          echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
               lib/ollama/cuda_v*)        echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
               lib/ollama/vulkan*)        echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
+              lib/ollama/mlx*)           echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}.tar.in ;;
               lib/ollama/cuda_jetpack5)  echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}-jetpack5.tar.in ;;
               lib/ollama/cuda_jetpack6)  echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}-jetpack6.tar.in ;;
               lib/ollama/rocm)           echo $COMPONENT >>ollama-${{ matrix.os }}-${{ matrix.arch }}-rocm.tar.in ;;

CMakeLists.txt

@@ -48,9 +48,10 @@ if((CMAKE_OSX_ARCHITECTURES AND NOT CMAKE_OSX_ARCHITECTURES MATCHES "arm64")
     set(GGML_CPU_ALL_VARIANTS ON)
 endif()
 
-if (CMAKE_OSX_ARCHITECTURES MATCHES "x86_64")
+if(APPLE)
     set(CMAKE_BUILD_RPATH "@loader_path")
     set(CMAKE_INSTALL_RPATH "@loader_path")
+    set(CMAKE_BUILD_WITH_INSTALL_RPATH ON)
 endif()
 
 set(OLLAMA_BUILD_DIR ${CMAKE_BINARY_DIR}/lib/ollama)
@@ -196,6 +197,14 @@ if(MLX_ENGINE)
         FRAMEWORK DESTINATION ${OLLAMA_INSTALL_DIR} COMPONENT MLX
     )
 
+    # Install the Metal library for macOS arm64 (must be colocated with the binary)
+    # Metal backend is only built for arm64, not x86_64
+    if(APPLE AND CMAKE_SYSTEM_PROCESSOR STREQUAL "arm64")
+        install(FILES ${CMAKE_BINARY_DIR}/_deps/mlx-build/mlx/backend/metal/kernels/mlx.metallib
+            DESTINATION ${OLLAMA_INSTALL_DIR}
+            COMPONENT MLX)
+    endif()
+
     # Manually install cudart and cublas since they might not be picked up as direct dependencies
     if(CUDAToolkit_FOUND)
         file(GLOB CUDART_LIBS

Dockerfile

@@ -161,6 +161,9 @@ ARG GOFLAGS="'-ldflags=-w -s'"
 ENV CGO_ENABLED=1
 ARG CGO_CFLAGS
 ARG CGO_CXXFLAGS
+RUN mkdir -p dist/bin
+RUN --mount=type=cache,target=/root/.cache/go-build \
+    go build -tags mlx -trimpath -buildmode=pie -o dist/bin/ollama-mlx .
 
 FROM base AS build
 WORKDIR /go/src/github.com/ollama/ollama
@@ -182,6 +185,7 @@ COPY --from=cuda-12 dist/lib/ollama /lib/ollama/
 COPY --from=cuda-13 dist/lib/ollama /lib/ollama/
 COPY --from=vulkan  dist/lib/ollama  /lib/ollama/
 COPY --from=mlx     /go/src/github.com/ollama/ollama/dist/lib/ollama /lib/ollama/
+COPY --from=mlx     /go/src/github.com/ollama/ollama/dist/bin/ /bin/
 
 FROM --platform=linux/arm64 scratch AS arm64
 # COPY --from=cuda-11 dist/lib/ollama/ /lib/ollama/

cmd/apps.go

@@ -1,558 +0,0 @@
-package cmd
-
-import (
-	"context"
-	"encoding/json"
-	"fmt"
-	"os"
-	"os/exec"
-	"path/filepath"
-	"strings"
-
-	"github.com/ollama/ollama/api"
-	"github.com/spf13/cobra"
-)
-
-type EnvVar struct {
-	Name  string
-	Value string
-}
-
-type AppConfig struct {
-	Name         string
-	DisplayName  string
-	Command      string
-	EnvVars      func(model string) []EnvVar
-	Args         func(model string) []string
-	Setup        func(model string) error
-	CheckInstall func() error
-}
-
-var ClaudeConfig = &AppConfig{
-	Name:        "Claude",
-	DisplayName: "Claude Code",
-	Command:     "claude",
-	EnvVars: func(model string) []EnvVar {
-		return []EnvVar{
-			{Name: "ANTHROPIC_BASE_URL", Value: "http://localhost:11434"},
-			{Name: "ANTHROPIC_API_KEY", Value: "ollama"},
-			{Name: "ANTHROPIC_AUTH_TOKEN", Value: "ollama"},
-		}
-	},
-	Args: func(model string) []string {
-		if model == "" {
-			return nil
-		}
-		return []string{"--model", model}
-	},
-	CheckInstall: func() error {
-		if _, err := exec.LookPath("claude"); err != nil {
-			return fmt.Errorf("claude is not installed. Install with: npm install -g @anthropic-ai/claude-code")
-		}
-		return nil
-	},
-}
-
-var DroidConfig = &AppConfig{
-	Name:        "Droid",
-	DisplayName: "Droid",
-	Command:     "droid",
-	EnvVars:     func(model string) []EnvVar { return nil },
-	Args:        func(model string) []string { return nil },
-	Setup:       setupDroidSettings,
-	CheckInstall: func() error {
-		if _, err := exec.LookPath("droid"); err != nil {
-			return fmt.Errorf("droid is not installed. Install from: https://docs.factory.ai/cli/install")
-		}
-		return nil
-	},
-}
-
-var AppRegistry = map[string]*AppConfig{
-	"claude":      ClaudeConfig,
-	"claude-code": ClaudeConfig,
-	"droid":       DroidConfig,
-	"opencode":    OpenCodeConfig,
-}
-
-func GetApp(name string) (*AppConfig, bool) {
-	app, ok := AppRegistry[strings.ToLower(name)]
-	return app, ok
-}
-
-func getModelContextLength(model string) int {
-	client, err := api.ClientFromEnvironment()
-	if err != nil {
-		return 8192
-	}
-	resp, err := client.Show(context.Background(), &api.ShowRequest{Model: model})
-	if err != nil || resp.ModelInfo == nil {
-		return 8192
-	}
-	arch, ok := resp.ModelInfo["general.architecture"].(string)
-	if !ok {
-		return 8192
-	}
-	if v, ok := resp.ModelInfo[fmt.Sprintf("%s.context_length", arch)].(float64); ok {
-		return min(int(v), 128000)
-	}
-	return 8192
-}
-
-func setupDroidSettings(model string) error {
-	home, err := os.UserHomeDir()
-	if err != nil {
-		return err
-	}
-
-	settingsPath := filepath.Join(home, ".factory", "settings.json")
-	if err := os.MkdirAll(filepath.Dir(settingsPath), 0755); err != nil {
-		return err
-	}
-
-	settings := make(map[string]any)
-	if data, err := os.ReadFile(settingsPath); err == nil {
-		json.Unmarshal(data, &settings)
-	}
-
-	var customModels []any
-	if existing, ok := settings["customModels"].([]any); ok {
-		customModels = existing
-	}
-
-	maxIndex := 0
-	existingIdx := -1
-	var existingID string
-	for i, m := range customModels {
-		if entry, ok := m.(map[string]any); ok {
-			if entry["model"] == model {
-				existingIdx = i
-				if id, ok := entry["id"].(string); ok {
-					existingID = id
-				}
-			}
-			if idx, ok := entry["index"].(float64); ok && int(idx) > maxIndex {
-				maxIndex = int(idx)
-			}
-		}
-	}
-
-	var modelID string
-	newEntry := map[string]any{
-		"model":           model,
-		"displayName":     fmt.Sprintf("%s [Ollama]", model),
-		"baseUrl":         "http://localhost:11434/v1",
-		"apiKey":          "ollama",
-		"provider":        "generic-chat-completion-api",
-		"maxOutputTokens": getModelContextLength(model),
-		"noImageSupport":  true,
-	}
-
-	if existingIdx >= 0 {
-		modelID = existingID
-		newEntry["id"] = existingID
-		customModels[existingIdx] = newEntry
-	} else {
-		newIndex := maxIndex + 1
-		modelID = fmt.Sprintf("custom:%s-[Ollama]-%d", model, newIndex)
-		newEntry["id"] = modelID
-		newEntry["index"] = newIndex
-		customModels = append(customModels, newEntry)
-	}
-	settings["customModels"] = customModels
-
-	sessionSettings, ok := settings["sessionDefaultSettings"].(map[string]any)
-	if !ok {
-		sessionSettings = make(map[string]any)
-	}
-	sessionSettings["model"] = modelID
-	settings["sessionDefaultSettings"] = sessionSettings
-
-	data, err := json.MarshalIndent(settings, "", "  ")
-	if err != nil {
-		return err
-	}
-	return os.WriteFile(settingsPath, data, 0644)
-}
-
-func setupOpenCodeSettings(model string) error {
-	home, err := os.UserHomeDir()
-	if err != nil {
-		return err
-	}
-
-	configPath := filepath.Join(home, ".config", "opencode", "opencode.json")
-	if err := os.MkdirAll(filepath.Dir(configPath), 0755); err != nil {
-		return err
-	}
-
-	config := make(map[string]any)
-	if data, err := os.ReadFile(configPath); err == nil {
-		json.Unmarshal(data, &config)
-	}
-
-	config["$schema"] = "https://opencode.ai/config.json"
-
-	provider, ok := config["provider"].(map[string]any)
-	if !ok {
-		provider = make(map[string]any)
-	}
-
-	ollama, ok := provider["ollama"].(map[string]any)
-	if !ok {
-		ollama = map[string]any{
-			"npm":  "@ai-sdk/openai-compatible",
-			"name": "Ollama (local)",
-			"options": map[string]any{
-				"baseURL": "http://localhost:11434/v1",
-			},
-		}
-	}
-
-	models, ok := ollama["models"].(map[string]any)
-	if !ok {
-		models = make(map[string]any)
-	}
-
-	models[model] = map[string]any{
-		"name": fmt.Sprintf("%s [Ollama]", model),
-	}
-
-	ollama["models"] = models
-	provider["ollama"] = ollama
-	config["provider"] = provider
-
-	data, err := json.MarshalIndent(config, "", "  ")
-	if err != nil {
-		return err
-	}
-	if err := os.WriteFile(configPath, data, 0644); err != nil {
-		return err
-	}
-
-	statePath := filepath.Join(home, ".local", "state", "opencode", "model.json")
-	if err := os.MkdirAll(filepath.Dir(statePath), 0755); err != nil {
-		return err
-	}
-
-	state := map[string]any{
-		"recent":   []any{},
-		"favorite": []any{},
-		"variant":  map[string]any{},
-	}
-	if data, err := os.ReadFile(statePath); err == nil {
-		json.Unmarshal(data, &state)
-	}
-
-	recent, _ := state["recent"].([]any)
-
-	newRecent := []any{}
-	for _, entry := range recent {
-		if e, ok := entry.(map[string]any); ok {
-			if e["providerID"] == "ollama" && e["modelID"] == model {
-				continue
-			}
-		}
-		newRecent = append(newRecent, entry)
-	}
-
-	newRecent = append([]any{map[string]any{
-		"providerID": "ollama",
-		"modelID":    model,
-	}}, newRecent...)
-
-	if len(newRecent) > 10 {
-		newRecent = newRecent[:10]
-	}
-
-	state["recent"] = newRecent
-
-	data, err = json.MarshalIndent(state, "", "  ")
-	if err != nil {
-		return err
-	}
-	return os.WriteFile(statePath, data, 0644)
-}
-
-var OpenCodeConfig = &AppConfig{
-	Name:        "OpenCode",
-	DisplayName: "OpenCode",
-	Command:     "opencode",
-	EnvVars:     func(model string) []EnvVar { return nil },
-	Args:        func(model string) []string { return nil },
-	Setup:       setupOpenCodeSettings,
-	CheckInstall: func() error {
-		if _, err := exec.LookPath("opencode"); err != nil {
-			return fmt.Errorf("opencode is not installed. Install from: https://opencode.ai")
-		}
-		return nil
-	},
-}
-
-func getOpenCodeConfiguredModel() string {
-	home, err := os.UserHomeDir()
-	if err != nil {
-		return ""
-	}
-
-	statePath := filepath.Join(home, ".local", "state", "opencode", "model.json")
-	data, err := os.ReadFile(statePath)
-	if err != nil {
-		return ""
-	}
-
-	var state map[string]any
-	if err := json.Unmarshal(data, &state); err != nil {
-		return ""
-	}
-
-	recent, ok := state["recent"].([]any)
-	if !ok || len(recent) == 0 {
-		return ""
-	}
-
-	first, ok := recent[0].(map[string]any)
-	if !ok {
-		return ""
-	}
-
-	if first["providerID"] == "ollama" {
-		if modelID, ok := first["modelID"].(string); ok {
-			return modelID
-		}
-	}
-	return ""
-}
-
-func getDroidConfiguredModel() string {
-	home, err := os.UserHomeDir()
-	if err != nil {
-		return ""
-	}
-
-	settingsPath := filepath.Join(home, ".factory", "settings.json")
-	data, err := os.ReadFile(settingsPath)
-	if err != nil {
-		return ""
-	}
-
-	var settings map[string]any
-	if err := json.Unmarshal(data, &settings); err != nil {
-		return ""
-	}
-
-	sessionSettings, ok := settings["sessionDefaultSettings"].(map[string]any)
-	if !ok {
-		return ""
-	}
-
-	modelID, ok := sessionSettings["model"].(string)
-	if !ok || modelID == "" {
-		return ""
-	}
-
-	customModels, ok := settings["customModels"].([]any)
-	if !ok {
-		return ""
-	}
-
-	for _, m := range customModels {
-		entry, ok := m.(map[string]any)
-		if !ok {
-			continue
-		}
-		if entry["id"] == modelID {
-			if model, ok := entry["model"].(string); ok {
-				return model
-			}
-		}
-	}
-	return ""
-}
-
-func runInApp(appName, modelName string) error {
-	app, ok := GetApp(appName)
-	if !ok {
-		return fmt.Errorf("unknown app: %s", appName)
-	}
-
-	if err := app.CheckInstall(); err != nil {
-		return err
-	}
-
-	if app.Setup != nil {
-		if err := app.Setup(modelName); err != nil {
-			return fmt.Errorf("setup failed: %w", err)
-		}
-	}
-
-	proc := exec.Command(app.Command, app.Args(modelName)...)
-	proc.Stdin = os.Stdin
-	proc.Stdout = os.Stdout
-	proc.Stderr = os.Stderr
-	proc.Env = os.Environ()
-	for _, env := range app.EnvVars(modelName) {
-		proc.Env = append(proc.Env, fmt.Sprintf("%s=%s", env.Name, env.Value))
-	}
-
-	fmt.Fprintf(os.Stderr, "Launching %s with %s...\n", app.DisplayName, modelName)
-	return proc.Run()
-}
-
-func ConnectCmd() *cobra.Command {
-	var modelName string
-
-	cmd := &cobra.Command{
-		Use:   "connect [APP]",
-		Short: "Configure an external app to use Ollama",
-		Long: `Configure an external application to use Ollama as its backend.
-
-Supported apps:
-  claude    Claude Code
-  droid     Droid
-  opencode  OpenCode
-
-Examples:
-  ollama connect
-  ollama connect claude
-  ollama connect claude --model llama3.2`,
-		Args:    cobra.MaximumNArgs(1),
-		PreRunE: checkServerHeartbeat,
-		RunE: func(cmd *cobra.Command, args []string) error {
-			var appName string
-			if len(args) > 0 {
-				appName = args[0]
-			} else {
-				var err error
-				appName, err = selectApp()
-				if err != nil {
-					return err
-				}
-			}
-
-			if _, ok := GetApp(appName); !ok {
-				return fmt.Errorf("unknown app: %s", appName)
-			}
-
-			if modelName == "" {
-				var err error
-				modelName, err = selectModelForConnect(cmd.Context(), "")
-				if err != nil {
-					return err
-				}
-			}
-
-			if err := SaveConnection(appName, modelName); err != nil {
-				return fmt.Errorf("failed to save: %w", err)
-			}
-
-			fmt.Fprintf(os.Stderr, "Added %s to %s\n", modelName, appName)
-
-			if launch, _ := confirmLaunch(appName); launch {
-				return runInApp(appName, modelName)
-			}
-
-			fmt.Fprintf(os.Stderr, "Run 'ollama launch %s' to start later\n", strings.ToLower(appName))
-			return nil
-		},
-	}
-
-	cmd.Flags().StringVar(&modelName, "model", "", "Model to use")
-	return cmd
-}
-
-func getAppConfiguredModel(appName string) string {
-	switch strings.ToLower(appName) {
-	case "opencode":
-		return getOpenCodeConfiguredModel()
-	case "droid":
-		return getDroidConfiguredModel()
-	default:
-		return ""
-	}
-}
-
-func LaunchCmd() *cobra.Command {
-	cmd := &cobra.Command{
-		Use:   "launch [APP]",
-		Short: "Launch a configured app",
-		Long: `Launch a configured application with Ollama as its backend.
-
-If no app is specified, shows a list of configured apps to choose from.
-If no apps have been configured, starts the connect flow.
-
-Examples:
-  ollama launch
-  ollama launch claude`,
-		Args:    cobra.MaximumNArgs(1),
-		PreRunE: checkServerHeartbeat,
-		RunE: func(cmd *cobra.Command, args []string) error {
-			var appName string
-			if len(args) > 0 {
-				appName = args[0]
-			} else {
-				selected, err := selectConnectedApp()
-				if err != nil {
-					return err
-				}
-				if selected == "" {
-					// No connected apps, start connect flow
-					fmt.Fprintf(os.Stderr, "No apps configured. Let's set one up.\n\n")
-					appName, err = selectApp()
-					if err != nil {
-						return err
-					}
-
-					modelName, err := selectModelForConnect(cmd.Context(), "")
-					if err != nil {
-						return err
-					}
-
-					if err := SaveConnection(appName, modelName); err != nil {
-						return fmt.Errorf("failed to save: %w", err)
-					}
-
-					fmt.Fprintf(os.Stderr, "Added %s to %s\n", modelName, appName)
-					return runInApp(appName, modelName)
-				}
-				appName = selected
-			}
-
-			app, ok := GetApp(appName)
-			if !ok {
-				return fmt.Errorf("unknown app: %s", appName)
-			}
-
-			// Check app's own config first
-			modelName := getAppConfiguredModel(appName)
-
-			// Fall back to our saved connection config
-			if modelName == "" {
-				config, err := LoadConnection(appName)
-				if err != nil {
-					if os.IsNotExist(err) {
-						// No config, drop into connect flow
-						modelName, err = selectModelForConnect(cmd.Context(), "")
-						if err != nil {
-							return err
-						}
-
-						if err := SaveConnection(appName, modelName); err != nil {
-							return fmt.Errorf("failed to save: %w", err)
-						}
-
-						fmt.Fprintf(os.Stderr, "Added %s to %s\n", modelName, appName)
-					} else {
-						return err
-					}
-				} else {
-					modelName = config.Model
-				}
-			}
-
-			return runInApp(app.Name, modelName)
-		},
-	}
-
-	return cmd
-}

cmd/apps_test.go

@@ -1,189 +0,0 @@
-package cmd
-
-import (
-	"encoding/json"
-	"os"
-	"path/filepath"
-	"testing"
-)
-
-func TestSetupOpenCodeSettings(t *testing.T) {
-	tmpDir := t.TempDir()
-	origHome := os.Getenv("HOME")
-	os.Setenv("HOME", tmpDir)
-	defer os.Setenv("HOME", origHome)
-
-	configDir := filepath.Join(tmpDir, ".config", "opencode")
-	configPath := filepath.Join(configDir, "opencode.json")
-	stateDir := filepath.Join(tmpDir, ".local", "state", "opencode")
-	statePath := filepath.Join(stateDir, "model.json")
-
-	cleanup := func() {
-		os.RemoveAll(configDir)
-		os.RemoveAll(stateDir)
-	}
-
-	t.Run("fresh install", func(t *testing.T) {
-		cleanup()
-		if err := setupOpenCodeSettings("llama3.2"); err != nil {
-			t.Fatal(err)
-		}
-		assertOpenCodeModelExists(t, configPath, "llama3.2")
-		assertOpenCodeRecentModel(t, statePath, 0, "ollama", "llama3.2")
-	})
-
-	t.Run("preserve other providers", func(t *testing.T) {
-		cleanup()
-		os.MkdirAll(configDir, 0755)
-		os.WriteFile(configPath, []byte(`{"provider":{"anthropic":{"apiKey":"xxx"}}}`), 0644)
-		if err := setupOpenCodeSettings("llama3.2"); err != nil {
-			t.Fatal(err)
-		}
-		data, _ := os.ReadFile(configPath)
-		var cfg map[string]any
-		json.Unmarshal(data, &cfg)
-		provider := cfg["provider"].(map[string]any)
-		if provider["anthropic"] == nil {
-			t.Error("anthropic provider was removed")
-		}
-		assertOpenCodeModelExists(t, configPath, "llama3.2")
-	})
-
-	t.Run("preserve other models", func(t *testing.T) {
-		cleanup()
-		os.MkdirAll(configDir, 0755)
-		os.WriteFile(configPath, []byte(`{"provider":{"ollama":{"models":{"mistral":{"name":"Mistral"}}}}}`), 0644)
-		if err := setupOpenCodeSettings("llama3.2"); err != nil {
-			t.Fatal(err)
-		}
-		assertOpenCodeModelExists(t, configPath, "mistral")
-		assertOpenCodeModelExists(t, configPath, "llama3.2")
-	})
-
-	t.Run("update existing model", func(t *testing.T) {
-		cleanup()
-		setupOpenCodeSettings("llama3.2")
-		setupOpenCodeSettings("llama3.2")
-		assertOpenCodeModelExists(t, configPath, "llama3.2")
-	})
-
-	t.Run("preserve top-level keys", func(t *testing.T) {
-		cleanup()
-		os.MkdirAll(configDir, 0755)
-		os.WriteFile(configPath, []byte(`{"theme":"dark","keybindings":{}}`), 0644)
-		if err := setupOpenCodeSettings("llama3.2"); err != nil {
-			t.Fatal(err)
-		}
-		data, _ := os.ReadFile(configPath)
-		var cfg map[string]any
-		json.Unmarshal(data, &cfg)
-		if cfg["theme"] != "dark" {
-			t.Error("theme was removed")
-		}
-		if cfg["keybindings"] == nil {
-			t.Error("keybindings was removed")
-		}
-	})
-
-	t.Run("model state - insert at index 0", func(t *testing.T) {
-		cleanup()
-		os.MkdirAll(stateDir, 0755)
-		os.WriteFile(statePath, []byte(`{"recent":[{"providerID":"anthropic","modelID":"claude"}],"favorite":[],"variant":{}}`), 0644)
-		if err := setupOpenCodeSettings("llama3.2"); err != nil {
-			t.Fatal(err)
-		}
-		assertOpenCodeRecentModel(t, statePath, 0, "ollama", "llama3.2")
-		assertOpenCodeRecentModel(t, statePath, 1, "anthropic", "claude")
-	})
-
-	t.Run("model state - preserve favorites and variants", func(t *testing.T) {
-		cleanup()
-		os.MkdirAll(stateDir, 0755)
-		os.WriteFile(statePath, []byte(`{"recent":[],"favorite":[{"providerID":"x","modelID":"y"}],"variant":{"a":"b"}}`), 0644)
-		if err := setupOpenCodeSettings("llama3.2"); err != nil {
-			t.Fatal(err)
-		}
-		data, _ := os.ReadFile(statePath)
-		var state map[string]any
-		json.Unmarshal(data, &state)
-		if len(state["favorite"].([]any)) != 1 {
-			t.Error("favorite was modified")
-		}
-		if state["variant"].(map[string]any)["a"] != "b" {
-			t.Error("variant was modified")
-		}
-	})
-
-	t.Run("model state - deduplicate on re-add", func(t *testing.T) {
-		cleanup()
-		os.MkdirAll(stateDir, 0755)
-		os.WriteFile(statePath, []byte(`{"recent":[{"providerID":"ollama","modelID":"llama3.2"},{"providerID":"anthropic","modelID":"claude"}],"favorite":[],"variant":{}}`), 0644)
-		if err := setupOpenCodeSettings("llama3.2"); err != nil {
-			t.Fatal(err)
-		}
-		data, _ := os.ReadFile(statePath)
-		var state map[string]any
-		json.Unmarshal(data, &state)
-		recent := state["recent"].([]any)
-		if len(recent) != 2 {
-			t.Errorf("expected 2 recent entries, got %d", len(recent))
-		}
-		assertOpenCodeRecentModel(t, statePath, 0, "ollama", "llama3.2")
-	})
-}
-
-func assertOpenCodeModelExists(t *testing.T, path, model string) {
-	t.Helper()
-	data, err := os.ReadFile(path)
-	if err != nil {
-		t.Fatal(err)
-	}
-	var cfg map[string]any
-	if err := json.Unmarshal(data, &cfg); err != nil {
-		t.Fatal(err)
-	}
-	provider, ok := cfg["provider"].(map[string]any)
-	if !ok {
-		t.Fatal("provider not found")
-	}
-	ollama, ok := provider["ollama"].(map[string]any)
-	if !ok {
-		t.Fatal("ollama provider not found")
-	}
-	models, ok := ollama["models"].(map[string]any)
-	if !ok {
-		t.Fatal("models not found")
-	}
-	if models[model] == nil {
-		t.Errorf("model %s not found", model)
-	}
-}
-
-func assertOpenCodeRecentModel(t *testing.T, path string, index int, providerID, modelID string) {
-	t.Helper()
-	data, err := os.ReadFile(path)
-	if err != nil {
-		t.Fatal(err)
-	}
-	var state map[string]any
-	if err := json.Unmarshal(data, &state); err != nil {
-		t.Fatal(err)
-	}
-	recent, ok := state["recent"].([]any)
-	if !ok {
-		t.Fatal("recent not found")
-	}
-	if index >= len(recent) {
-		t.Fatalf("index %d out of range (len=%d)", index, len(recent))
-	}
-	entry, ok := recent[index].(map[string]any)
-	if !ok {
-		t.Fatal("entry is not a map")
-	}
-	if entry["providerID"] != providerID {
-		t.Errorf("expected providerID %s, got %s", providerID, entry["providerID"])
-	}
-	if entry["modelID"] != modelID {
-		t.Errorf("expected modelID %s, got %s", modelID, entry["modelID"])
-	}
-}

cmd/cmd.go

@@ -1945,8 +1945,6 @@ func NewCLI() *cobra.Command {
 		copyCmd,
 		deleteCmd,
 		runnerCmd,
-		ConnectCmd(),
-		LaunchCmd(),
 	)
 
 	return rootCmd

cmd/config.go

@@ -1,99 +0,0 @@
-package cmd
-
-import (
-	"encoding/json"
-	"os"
-	"path/filepath"
-	"time"
-)
-
-type ConnectionConfig struct {
-	App          string    `json:"app"`
-	Model        string    `json:"model"`
-	ConfiguredAt time.Time `json:"configured_at"`
-}
-
-func configPath(appName string) (string, error) {
-	home, err := os.UserHomeDir()
-	if err != nil {
-		return "", err
-	}
-	return filepath.Join(home, ".ollama", "connections", appName+".json"), nil
-}
-
-func SaveConnection(appName, model string) error {
-	path, err := configPath(appName)
-	if err != nil {
-		return err
-	}
-
-	if err := os.MkdirAll(filepath.Dir(path), 0755); err != nil {
-		return err
-	}
-
-	data, err := json.MarshalIndent(ConnectionConfig{
-		App:          appName,
-		Model:        model,
-		ConfiguredAt: time.Now(),
-	}, "", "  ")
-	if err != nil {
-		return err
-	}
-
-	return os.WriteFile(path, data, 0644)
-}
-
-func LoadConnection(appName string) (*ConnectionConfig, error) {
-	path, err := configPath(appName)
-	if err != nil {
-		return nil, err
-	}
-
-	data, err := os.ReadFile(path)
-	if err != nil {
-		return nil, err
-	}
-
-	var config ConnectionConfig
-	if err := json.Unmarshal(data, &config); err != nil {
-		return nil, err
-	}
-
-	return &config, nil
-}
-
-func ListConnections() ([]ConnectionConfig, error) {
-	home, err := os.UserHomeDir()
-	if err != nil {
-		return nil, err
-	}
-
-	dir := filepath.Join(home, ".ollama", "connections")
-	entries, err := os.ReadDir(dir)
-	if err != nil {
-		if os.IsNotExist(err) {
-			return nil, nil
-		}
-		return nil, err
-	}
-
-	var configs []ConnectionConfig
-	for _, entry := range entries {
-		if entry.IsDir() || filepath.Ext(entry.Name()) != ".json" {
-			continue
-		}
-
-		data, err := os.ReadFile(filepath.Join(dir, entry.Name()))
-		if err != nil {
-			continue
-		}
-
-		var config ConnectionConfig
-		if err := json.Unmarshal(data, &config); err != nil {
-			continue
-		}
-		configs = append(configs, config)
-	}
-
-	return configs, nil
-}

cmd/interactive.go

@@ -37,8 +37,6 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 		fmt.Fprintln(os.Stderr, "  /load <model>   Load a session or model")
 		fmt.Fprintln(os.Stderr, "  /save <model>   Save your current session")
 		fmt.Fprintln(os.Stderr, "  /clear          Clear session context")
-		fmt.Fprintln(os.Stderr, "  /connect        Configure an external app to use Ollama")
-		fmt.Fprintln(os.Stderr, "  /launch [app]   Launch a configured app")
 		fmt.Fprintln(os.Stderr, "  /bye            Exit")
 		fmt.Fprintln(os.Stderr, "  /?, /help       Help for a command")
 		fmt.Fprintln(os.Stderr, "  /? shortcuts    Help for keyboard shortcuts")
@@ -462,136 +460,6 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 			}
 		case strings.HasPrefix(line, "/exit"), strings.HasPrefix(line, "/bye"):
 			return nil
-		case strings.HasPrefix(line, "/connect"):
-			args := strings.Fields(line)
-			var appName string
-			if len(args) > 1 {
-				appName = args[1]
-			} else {
-				var err error
-				appName, err = selectApp()
-				if err != nil {
-					fmt.Printf("error: %v\n", err)
-					continue
-				}
-			}
-
-			if _, ok := GetApp(appName); !ok {
-				fmt.Printf("Unknown app: %s\n", appName)
-				continue
-			}
-
-			modelName, err := selectModelForConnect(cmd.Context(), opts.Model)
-			if err != nil {
-				fmt.Printf("error: %v\n", err)
-				continue
-			}
-
-			if err := SaveConnection(appName, modelName); err != nil {
-				fmt.Printf("error: %v\n", err)
-				continue
-			}
-
-			fmt.Fprintf(os.Stderr, "Added %s to %s\n", modelName, appName)
-
-			if launch, _ := confirmLaunch(appName); launch {
-				if err := runInApp(appName, modelName); err != nil {
-					fmt.Printf("error: %v\n", err)
-				}
-			}
-			continue
-		case strings.HasPrefix(line, "/launch"):
-			args := strings.Fields(line)
-			var appName string
-			if len(args) >= 2 {
-				appName = args[1]
-			} else {
-				selected, err := selectConnectedApp()
-				if err != nil {
-					fmt.Printf("error: %v\n", err)
-					continue
-				}
-				if selected == "" {
-					// No connected apps, start connect flow
-					fmt.Fprintf(os.Stderr, "No apps configured. Let's set one up.\n\n")
-					appName, err = selectApp()
-					if err != nil {
-						fmt.Printf("error: %v\n", err)
-						continue
-					}
-
-					modelName, err := selectModelForConnect(cmd.Context(), opts.Model)
-					if err != nil {
-						fmt.Printf("error: %v\n", err)
-						continue
-					}
-
-					if err := SaveConnection(appName, modelName); err != nil {
-						fmt.Printf("error: %v\n", err)
-						continue
-					}
-
-					fmt.Fprintf(os.Stderr, "Added %s to %s\n", modelName, appName)
-					if err := runInApp(appName, modelName); err != nil {
-						fmt.Printf("error: %v\n", err)
-					}
-					continue
-				}
-				appName = selected
-			}
-
-			app, ok := GetApp(appName)
-			if !ok {
-				fmt.Printf("Unknown app: %s\n", appName)
-				continue
-			}
-
-			// Check app's own config first
-			modelName := getAppConfiguredModel(appName)
-
-			// Fall back to our saved connection config
-			if modelName == "" {
-				config, err := LoadConnection(appName)
-				if err != nil {
-					if os.IsNotExist(err) {
-						// No config, drop into connect flow
-						modelName, err = selectModelForConnect(cmd.Context(), opts.Model)
-						if err != nil {
-							fmt.Printf("error: %v\n", err)
-							continue
-						}
-
-						if err := SaveConnection(appName, modelName); err != nil {
-							fmt.Printf("error: %v\n", err)
-							continue
-						}
-
-						fmt.Fprintf(os.Stderr, "Added %s to %s\n", modelName, appName)
-					} else {
-						fmt.Printf("error: %v\n", err)
-						continue
-					}
-				} else {
-					modelName = config.Model
-				}
-			}
-
-			// If running model differs from configured, offer to switch
-			if opts.Model != "" && modelName != opts.Model {
-				if switchModel, _ := confirmPrompt(fmt.Sprintf("Switch %s to use %s?", app.DisplayName, opts.Model)); switchModel {
-					modelName = opts.Model
-					if err := SaveConnection(appName, modelName); err != nil {
-						fmt.Printf("error: %v\n", err)
-						continue
-					}
-					fmt.Fprintf(os.Stderr, "Updated %s to %s\n", appName, modelName)
-				}
-			}
-
-			if err := runInApp(app.Name, modelName); err != nil {
-				fmt.Printf("error: %v\n", err)
-			}
-			continue
 		case strings.HasPrefix(line, "/"):
 			args := strings.Fields(line)
 			isFile := false

cmd/selector.go

@@ -1,328 +0,0 @@
-package cmd
-
-import (
-	"context"
-	"errors"
-	"fmt"
-	"os"
-	"sort"
-	"strings"
-	"time"
-
-	"golang.org/x/term"
-
-	"github.com/ollama/ollama/api"
-)
-
-const maxDisplayedItems = 10
-
-var AppOrder = []string{"claude", "opencode", "droid"}
-
-type SelectItem struct {
-	Name        string
-	Description string
-}
-
-func Select(prompt string, items []SelectItem) (string, error) {
-	if len(items) == 0 {
-		return "", fmt.Errorf("no items to select from")
-	}
-
-	fd := int(os.Stdin.Fd())
-	oldState, err := term.MakeRaw(fd)
-	if err != nil {
-		return "", err
-	}
-	defer term.Restore(fd, oldState)
-
-	fmt.Fprint(os.Stderr, "\033[?25l")
-	defer fmt.Fprint(os.Stderr, "\033[?25h")
-
-	var filter string
-	selected := 0
-	scrollOffset := 0
-	var lastLineCount int
-
-	render := func() {
-		filtered := filterItems(items, filter)
-
-		if lastLineCount > 0 {
-			fmt.Fprintf(os.Stderr, "\033[%dA", lastLineCount)
-		}
-		fmt.Fprint(os.Stderr, "\033[J")
-
-		fmt.Fprintf(os.Stderr, "%s %s\r\n", prompt, filter)
-		lineCount := 1
-
-		if len(filtered) == 0 {
-			fmt.Fprintf(os.Stderr, "  \033[37m(no matches)\033[0m\r\n")
-			lineCount++
-		} else {
-			displayCount := min(len(filtered), maxDisplayedItems)
-
-			for i := 0; i < displayCount; i++ {
-				idx := scrollOffset + i
-				if idx >= len(filtered) {
-					break
-				}
-				item := filtered[idx]
-				if idx == selected {
-					if item.Description != "" {
-						fmt.Fprintf(os.Stderr, "  \033[1m> %s\033[0m \033[37m- %s\033[0m\r\n", item.Name, item.Description)
-					} else {
-						fmt.Fprintf(os.Stderr, "  \033[1m> %s\033[0m\r\n", item.Name)
-					}
-				} else {
-					if item.Description != "" {
-						fmt.Fprintf(os.Stderr, "    %s \033[37m- %s\033[0m\r\n", item.Name, item.Description)
-					} else {
-						fmt.Fprintf(os.Stderr, "    %s\r\n", item.Name)
-					}
-				}
-				lineCount++
-			}
-
-			if remaining := len(filtered) - scrollOffset - displayCount; remaining > 0 {
-				fmt.Fprintf(os.Stderr, "  \033[37m... and %d more\033[0m\r\n", remaining)
-				lineCount++
-			}
-		}
-
-		lastLineCount = lineCount
-	}
-
-	clearUI := func() {
-		if lastLineCount > 0 {
-			fmt.Fprintf(os.Stderr, "\033[%dA", lastLineCount)
-			fmt.Fprint(os.Stderr, "\033[J")
-		}
-	}
-
-	render()
-
-	buf := make([]byte, 3)
-	for {
-		n, err := os.Stdin.Read(buf)
-		if err != nil {
-			return "", err
-		}
-
-		filtered := filterItems(items, filter)
-
-		switch {
-		case n == 1 && buf[0] == 13: // Enter
-			if len(filtered) > 0 && selected < len(filtered) {
-				clearUI()
-				return filtered[selected].Name, nil
-			}
-		case n == 1 && (buf[0] == 3 || buf[0] == 27): // Ctrl+C or Escape
-			clearUI()
-			return "", fmt.Errorf("cancelled")
-		case n == 1 && buf[0] == 127: // Backspace
-			if len(filter) > 0 {
-				filter = filter[:len(filter)-1]
-				selected = 0
-				scrollOffset = 0
-			}
-		case n == 3 && buf[0] == 27 && buf[1] == 91: // Arrow keys
-			if buf[2] == 65 && selected > 0 { // Up
-				selected--
-				if selected < scrollOffset {
-					scrollOffset = selected
-				}
-			} else if buf[2] == 66 && selected < len(filtered)-1 { // Down
-				selected++
-				if selected >= scrollOffset+maxDisplayedItems {
-					scrollOffset = selected - maxDisplayedItems + 1
-				}
-			}
-		case n == 1 && buf[0] >= 32 && buf[0] < 127: // Printable chars
-			filter += string(buf[0])
-			selected = 0
-			scrollOffset = 0
-		}
-
-		render()
-	}
-}
-
-func filterItems(items []SelectItem, filter string) []SelectItem {
-	if filter == "" {
-		return items
-	}
-	var result []SelectItem
-	filterLower := strings.ToLower(filter)
-	for _, item := range items {
-		if strings.Contains(strings.ToLower(item.Name), filterLower) {
-			result = append(result, item)
-		}
-	}
-	return result
-}
-
-func selectApp() (string, error) {
-	var items []SelectItem
-
-	for _, name := range AppOrder {
-		app, ok := AppRegistry[name]
-		if !ok {
-			continue
-		}
-		items = append(items, SelectItem{Name: app.Name, Description: app.DisplayName})
-	}
-
-	if len(items) == 0 {
-		return "", fmt.Errorf("no apps available")
-	}
-
-	return Select("Select app:", items)
-}
-
-func selectConnectedApp() (string, error) {
-	connections, err := ListConnections()
-	if err != nil {
-		return "", err
-	}
-
-	if len(connections) == 0 {
-		return "", nil
-	}
-
-	var items []SelectItem
-	for _, conn := range connections {
-		app, ok := GetApp(conn.App)
-		if !ok {
-			continue
-		}
-		items = append(items, SelectItem{
-			Name:        app.Name,
-			Description: fmt.Sprintf("%s (%s)", app.DisplayName, conn.Model),
-		})
-	}
-
-	if len(items) == 0 {
-		return "", nil
-	}
-
-	return Select("Select app to launch:", items)
-}
-
-func confirmLaunch(appName string) (bool, error) {
-	return confirmPrompt(fmt.Sprintf("Launch %s now?", appName))
-}
-
-func confirmPrompt(prompt string) (bool, error) {
-	fd := int(os.Stdin.Fd())
-	oldState, err := term.MakeRaw(fd)
-	if err != nil {
-		return false, err
-	}
-	defer term.Restore(fd, oldState)
-
-	fmt.Fprintf(os.Stderr, "%s [y/n] ", prompt)
-
-	buf := make([]byte, 1)
-	for {
-		if _, err := os.Stdin.Read(buf); err != nil {
-			return false, err
-		}
-
-		switch buf[0] {
-		case 'Y', 'y', 13:
-			fmt.Fprintf(os.Stderr, "yes\r\n")
-			return true, nil
-		case 'N', 'n', 27, 3:
-			fmt.Fprintf(os.Stderr, "no\r\n")
-			return false, nil
-		}
-	}
-}
-
-func selectModelForConnect(ctx context.Context, currentModel string) (string, error) {
-	client, err := api.ClientFromEnvironment()
-	if err != nil {
-		return "", err
-	}
-
-	models, err := client.List(ctx)
-	if err != nil {
-		return "", err
-	}
-
-	if len(models.Models) == 0 {
-		return "", fmt.Errorf("no models available. Run 'ollama pull <model>' first")
-	}
-
-	var items []SelectItem
-	cloudModels := make(map[string]bool)
-	for _, m := range models.Models {
-		items = append(items, SelectItem{Name: m.Name})
-		if m.RemoteModel != "" {
-			cloudModels[m.Name] = true
-		}
-	}
-
-	sort.Slice(items, func(i, j int) bool {
-		return strings.ToLower(items[i].Name) < strings.ToLower(items[j].Name)
-	})
-
-	if currentModel != "" {
-		for i, item := range items {
-			if item.Name == currentModel {
-				items = append([]SelectItem{item}, append(items[:i], items[i+1:]...)...)
-				break
-			}
-		}
-	}
-
-	selected, err := Select("Select model:", items)
-	if err != nil {
-		return "", err
-	}
-
-	if cloudModels[selected] {
-		if err := ensureSignedIn(ctx, client); err != nil {
-			return "", err
-		}
-	}
-
-	return selected, nil
-}
-
-func ensureSignedIn(ctx context.Context, client *api.Client) error {
-	user, err := client.Whoami(ctx)
-	if err == nil && user != nil && user.Name != "" {
-		return nil
-	}
-
-	var aErr api.AuthorizationError
-	if !errors.As(err, &aErr) || aErr.SigninURL == "" {
-		return err
-	}
-
-	yes, err := confirmPrompt("Sign in to ollama.com?")
-	if err != nil || !yes {
-		return fmt.Errorf("sign in required for cloud models")
-	}
-
-	fmt.Fprintf(os.Stderr, "\nTo sign in, navigate to:\n    %s\n\n", aErr.SigninURL)
-	fmt.Fprintf(os.Stderr, "\033[90mwaiting for sign in to complete...\033[0m")
-
-	ticker := time.NewTicker(2 * time.Second)
-	defer ticker.Stop()
-
-	for {
-		select {
-		case <-ctx.Done():
-			fmt.Fprintf(os.Stderr, "\n")
-			return ctx.Err()
-		case <-ticker.C:
-			user, err := client.Whoami(ctx)
-			if err == nil && user != nil && user.Name != "" {
-				fmt.Fprintf(os.Stderr, "\r\033[K\033[A\r\033[K\033[1msigned in:\033[0m %s\n", user.Name)
-				return nil
-			}
-			fmt.Fprintf(os.Stderr, ".")
-		}
-	}
-}

scripts/build_darwin.sh

@@ -73,7 +73,7 @@ _build_darwin() {
             MLX_CGO_CFLAGS="-O3 -I$(pwd)/$BUILD_DIR/_deps/mlx-c-src -mmacosx-version-min=14.0"
             MLX_CGO_LDFLAGS="-L$(pwd)/$BUILD_DIR/lib/ollama -lmlxc -lmlx -Wl,-rpath,@executable_path -lc++ -framework Metal -framework Foundation -framework Accelerate -mmacosx-version-min=14.0"
         fi
-        GOOS=darwin GOARCH=$ARCH CGO_ENABLED=1 CGO_CFLAGS="$MLX_CGO_CFLAGS" CGO_LDFLAGS="$MLX_CGO_LDFLAGS" go build -tags mlx -o $INSTALL_PREFIX/imagegen ./x/imagegen/cmd/engine
+        GOOS=darwin GOARCH=$ARCH CGO_ENABLED=1 CGO_CFLAGS="$MLX_CGO_CFLAGS" CGO_LDFLAGS="$MLX_CGO_LDFLAGS" go build -tags mlx -o $INSTALL_PREFIX/ollama-mlx .
         GOOS=darwin GOARCH=$ARCH CGO_ENABLED=1 go build -o $INSTALL_PREFIX .
     done
 }
@@ -82,19 +82,19 @@ _sign_darwin() {
     status "Creating universal binary..."
     mkdir -p dist/darwin
     lipo -create -output dist/darwin/ollama dist/darwin-*/ollama
-    lipo -create -output dist/darwin/imagegen dist/darwin-*/imagegen
+    lipo -create -output dist/darwin/ollama-mlx dist/darwin-*/ollama-mlx
     chmod +x dist/darwin/ollama
-    chmod +x dist/darwin/imagegen
+    chmod +x dist/darwin/ollama-mlx
 
     if [ -n "$APPLE_IDENTITY" ]; then
-        for F in dist/darwin/ollama dist/darwin-*/lib/ollama/* dist/darwin/imagegen; do
+        for F in dist/darwin/ollama dist/darwin-*/lib/ollama/* dist/darwin/ollama-mlx; do
             codesign -f --timestamp -s "$APPLE_IDENTITY" --identifier ai.ollama.ollama --options=runtime $F
         done
 
         # create a temporary zip for notarization
         TEMP=$(mktemp -u).zip
         ditto -c -k --keepParent dist/darwin/ollama "$TEMP"
-        xcrun notarytool submit "$TEMP" --wait --timeout 10m --apple-id $APPLE_ID --password $APPLE_PASSWORD --team-id $APPLE_TEAM_ID
+        xcrun notarytool submit "$TEMP" --wait --timeout 20m --apple-id $APPLE_ID --password $APPLE_PASSWORD --team-id $APPLE_TEAM_ID
         rm -f "$TEMP"
     fi
 
@@ -154,23 +154,25 @@ _build_macapp() {
     mkdir -p dist/Ollama.app/Contents/Resources
     if [ -d dist/darwin-amd64 ]; then
         lipo -create -output dist/Ollama.app/Contents/Resources/ollama dist/darwin-amd64/ollama dist/darwin-arm64/ollama
-        lipo -create -output dist/Ollama.app/Contents/Resources/imagegen dist/darwin-amd64/imagegen dist/darwin-arm64/imagegen
+        lipo -create -output dist/Ollama.app/Contents/Resources/ollama-mlx dist/darwin-amd64/ollama-mlx dist/darwin-arm64/ollama-mlx
         for F in dist/darwin-amd64/lib/ollama/*mlx*.dylib ; do
             lipo -create -output dist/darwin/$(basename $F) $F dist/darwin-arm64/lib/ollama/$(basename $F)
         done
         cp dist/darwin-*/lib/ollama/*.so dist/darwin-*/lib/ollama/*.dylib dist/Ollama.app/Contents/Resources/
         cp dist/darwin/*.dylib dist/Ollama.app/Contents/Resources/
+        # Copy MLX metallib (architecture-independent, just use arm64 version)
+        cp dist/darwin-arm64/lib/ollama/*.metallib dist/Ollama.app/Contents/Resources/ 2>/dev/null || true
     else
         cp -a dist/darwin/ollama dist/Ollama.app/Contents/Resources/ollama
         cp dist/darwin/*.so dist/darwin/*.dylib dist/Ollama.app/Contents/Resources/
     fi
-    cp -a dist/darwin/imagegen dist/Ollama.app/Contents/Resources/imagegen
+    cp -a dist/darwin/ollama-mlx dist/Ollama.app/Contents/Resources/ollama-mlx
     chmod a+x dist/Ollama.app/Contents/Resources/ollama
 
     # Sign
     if [ -n "$APPLE_IDENTITY" ]; then
         codesign -f --timestamp -s "$APPLE_IDENTITY" --identifier ai.ollama.ollama --options=runtime dist/Ollama.app/Contents/Resources/ollama
-        for lib in dist/Ollama.app/Contents/Resources/*.so dist/Ollama.app/Contents/Resources/*.dylib dist/Ollama.app/Contents/Resources/imagegen ; do
+        for lib in dist/Ollama.app/Contents/Resources/*.so dist/Ollama.app/Contents/Resources/*.dylib dist/Ollama.app/Contents/Resources/*.metallib dist/Ollama.app/Contents/Resources/ollama-mlx ; do
             codesign -f --timestamp -s "$APPLE_IDENTITY" --identifier ai.ollama.ollama --options=runtime ${lib}
         done
         codesign -f --timestamp -s "$APPLE_IDENTITY" --identifier com.electron.ollama --deep --options=runtime dist/Ollama.app
@@ -178,11 +180,11 @@ _build_macapp() {
 
     rm -f dist/Ollama-darwin.zip
     ditto -c -k --keepParent dist/Ollama.app dist/Ollama-darwin.zip
-    (cd dist/Ollama.app/Contents/Resources/; tar -cf - ollama imagegen *.so *.dylib) | gzip -9vc > dist/ollama-darwin.tgz
+    (cd dist/Ollama.app/Contents/Resources/; tar -cf - ollama ollama-mlx *.so *.dylib *.metallib 2>/dev/null) | gzip -9vc > dist/ollama-darwin.tgz
 
     # Notarize and Staple
     if [ -n "$APPLE_IDENTITY" ]; then
-        $(xcrun -f notarytool) submit dist/Ollama-darwin.zip --wait --timeout 10m --apple-id "$APPLE_ID" --password "$APPLE_PASSWORD" --team-id "$APPLE_TEAM_ID"
+        $(xcrun -f notarytool) submit dist/Ollama-darwin.zip --wait --timeout 20m --apple-id "$APPLE_ID" --password "$APPLE_PASSWORD" --team-id "$APPLE_TEAM_ID"
         rm -f dist/Ollama-darwin.zip
         $(xcrun -f stapler) staple dist/Ollama.app
         ditto -c -k --keepParent dist/Ollama.app dist/Ollama-darwin.zip
@@ -206,7 +208,7 @@ _build_macapp() {
         rm -f dist/rw*.dmg
 
         codesign -f --timestamp -s "$APPLE_IDENTITY" --identifier ai.ollama.ollama --options=runtime dist/Ollama.dmg
-        $(xcrun -f notarytool) submit dist/Ollama.dmg --wait --timeout 10m --apple-id "$APPLE_ID" --password "$APPLE_PASSWORD" --team-id "$APPLE_TEAM_ID"
+        $(xcrun -f notarytool) submit dist/Ollama.dmg --wait --timeout 20m --apple-id "$APPLE_ID" --password "$APPLE_PASSWORD" --team-id "$APPLE_TEAM_ID"
         $(xcrun -f stapler) staple dist/Ollama.dmg
     else
         echo "WARNING: Code signing disabled, this bundle will not work for upgrade testing"

scripts/build_linux.sh

@@ -48,53 +48,12 @@ if echo $PLATFORM | grep "amd64" > /dev/null; then
         .
 fi
 
-# Deduplicate CUDA libraries across mlx_* and cuda_* directories
-deduplicate_cuda_libs() {
-    local base_dir="$1"
-    echo "Deduplicating CUDA libraries in ${base_dir}..."
-
-    # Find all mlx_cuda_* directories
-    for mlx_dir in "${base_dir}"/lib/ollama/mlx_cuda_*; do
-        [ -d "${mlx_dir}" ] || continue
-
-        # Extract CUDA version (e.g., v12, v13)
-        cuda_version=$(basename "${mlx_dir}" | sed 's/mlx_cuda_//')
-        cuda_dir="${base_dir}/lib/ollama/cuda_${cuda_version}"
-
-        # Skip if corresponding cuda_* directory doesn't exist
-        [ -d "${cuda_dir}" ] || continue
-
-        echo "  Checking ${mlx_dir} against ${cuda_dir}..."
-
-        # Find all .so* files in mlx directory
-        find "${mlx_dir}" -type f -name "*.so*" | while read mlx_file; do
-            filename=$(basename "${mlx_file}")
-            cuda_file="${cuda_dir}/${filename}"
-
-            # Skip if file doesn't exist in cuda directory
-            [ -f "${cuda_file}" ] || continue
-
-            # Compare checksums
-            mlx_sum=$(sha256sum "${mlx_file}" | awk '{print $1}')
-            cuda_sum=$(sha256sum "${cuda_file}" | awk '{print $1}')
-
-            if [ "${mlx_sum}" = "${cuda_sum}" ]; then
-                echo "    Deduplicating ${filename}"
-                # Calculate relative path from mlx_dir to cuda_dir
-                rel_path="../cuda_${cuda_version}/${filename}"
-                rm -f "${mlx_file}"
-                ln -s "${rel_path}" "${mlx_file}"
-            fi
-        done
-    done
-}
-
 # Run deduplication for each platform output directory
 if echo $PLATFORM | grep "," > /dev/null ; then
-    deduplicate_cuda_libs "./dist/linux_amd64"
-    deduplicate_cuda_libs "./dist/linux_arm64"
+    $(dirname $0)/deduplicate_cuda_libs.sh "./dist/linux_amd64"
+    $(dirname $0)/deduplicate_cuda_libs.sh "./dist/linux_arm64"
 elif echo $PLATFORM | grep "amd64\|arm64" > /dev/null ; then
-    deduplicate_cuda_libs "./dist"
+    $(dirname $0)/deduplicate_cuda_libs.sh "./dist"
 fi
 
 # buildx behavior changes for single vs. multiplatform

scripts/deduplicate_cuda_libs.sh

@@ -0,0 +1,60 @@
+#!/bin/sh
+#
+# Deduplicate CUDA libraries across mlx_* and cuda_* directories
+# This script finds identical .so* files in mlx_cuda_* directories that exist
+# in corresponding cuda_* directories and replaces them with symlinks.
+#
+
+set -eu
+
+if [ $# -eq 0 ]; then
+    echo "ERROR: No directory specified" >&2
+    echo "Usage: $0 <base_directory>" >&2
+    exit 1
+fi
+
+base_dir="$1"
+
+if [ ! -d "${base_dir}" ]; then
+    echo "ERROR: Directory ${base_dir} does not exist" >&2
+    exit 1
+fi
+
+echo "Deduplicating CUDA libraries in ${base_dir}..."
+
+# Find all mlx_cuda_* directories
+for mlx_dir in "${base_dir}"/lib/ollama/mlx_cuda_*; do
+    [ -d "${mlx_dir}" ] || continue
+
+    # Extract CUDA version (e.g., v12, v13)
+    cuda_version=$(basename "${mlx_dir}" | sed 's/mlx_cuda_//')
+    cuda_dir="${base_dir}/lib/ollama/cuda_${cuda_version}"
+
+    # Skip if corresponding cuda_* directory doesn't exist
+    [ -d "${cuda_dir}" ] || continue
+
+    echo "  Checking ${mlx_dir} against ${cuda_dir}..."
+
+    # Find all .so* files in mlx directory
+    find "${mlx_dir}" -type f -name "*.so*" | while read mlx_file; do
+        filename=$(basename "${mlx_file}")
+        cuda_file="${cuda_dir}/${filename}"
+
+        # Skip if file doesn't exist in cuda directory
+        [ -f "${cuda_file}" ] || continue
+
+        # Compare checksums
+        mlx_sum=$(sha256sum "${mlx_file}" | awk '{print $1}')
+        cuda_sum=$(sha256sum "${cuda_file}" | awk '{print $1}')
+
+        if [ "${mlx_sum}" = "${cuda_sum}" ]; then
+            echo "    Deduplicating ${filename}"
+            # Calculate relative path from mlx_dir to cuda_dir
+            rel_path="../cuda_${cuda_version}/${filename}"
+            rm -f "${mlx_file}"
+            ln -s "${rel_path}" "${mlx_file}"
+        fi
+    done
+done
+
+echo "Deduplication complete"

server/download.go

@@ -95,11 +95,48 @@ func (p *blobDownloadPart) UnmarshalJSON(b []byte) error {
 }
 
 const (
-	numDownloadParts          = 16
+	// numDownloadParts is the default number of concurrent download parts for standard downloads
+	numDownloadParts = 16
+	// numHFDownloadParts is the reduced number of concurrent download parts for HuggingFace
+	// downloads to avoid triggering rate limits (HTTP 429 errors). See GitHub issue #13297.
+	numHFDownloadParts        = 4
 	minDownloadPartSize int64 = 100 * format.MegaByte
 	maxDownloadPartSize int64 = 1000 * format.MegaByte
 )
 
+// isHuggingFaceURL returns true if the URL is from a HuggingFace domain.
+// This includes:
+//   - huggingface.co (main domain)
+//   - *.huggingface.co (subdomains like cdn-lfs.huggingface.co)
+//   - hf.co (shortlink domain)
+//   - *.hf.co (CDN domains like cdn-lfs.hf.co, cdn-lfs3.hf.co)
+func isHuggingFaceURL(u *url.URL) bool {
+	if u == nil {
+		return false
+	}
+	host := strings.ToLower(u.Hostname())
+	return host == "huggingface.co" ||
+		strings.HasSuffix(host, ".huggingface.co") ||
+		host == "hf.co" ||
+		strings.HasSuffix(host, ".hf.co")
+}
+
+// getNumDownloadParts returns the number of concurrent download parts to use
+// for the given URL. HuggingFace URLs use reduced concurrency (default 4) to
+// avoid triggering rate limits. This can be overridden via the OLLAMA_HF_CONCURRENCY
+// environment variable. For non-HuggingFace URLs, returns the standard concurrency (16).
+func getNumDownloadParts(u *url.URL) int {
+	if isHuggingFaceURL(u) {
+		if v := os.Getenv("OLLAMA_HF_CONCURRENCY"); v != "" {
+			if n, err := strconv.Atoi(v); err == nil && n > 0 {
+				return n
+			}
+		}
+		return numHFDownloadParts
+	}
+	return numDownloadParts
+}
+
 func (p *blobDownloadPart) Name() string {
 	return strings.Join([]string{
 		p.blobDownload.Name, "partial", strconv.Itoa(p.N),
@@ -271,7 +308,11 @@ func (b *blobDownload) run(ctx context.Context, requestURL *url.URL, opts *regis
 	}
 
 	g, inner := errgroup.WithContext(ctx)
-	g.SetLimit(numDownloadParts)
+	concurrency := getNumDownloadParts(directURL)
+	if concurrency != numDownloadParts {
+		slog.Info(fmt.Sprintf("using reduced concurrency (%d) for HuggingFace download", concurrency))
+	}
+	g.SetLimit(concurrency)
 	for i := range b.Parts {
 		part := b.Parts[i]
 		if part.Completed.Load() == part.Size {

server/download_test.go

@@ -0,0 +1,194 @@
+package server
+
+import (
+	"net/url"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestIsHuggingFaceURL(t *testing.T) {
+	tests := []struct {
+		name     string
+		url      string
+		expected bool
+	}{
+		{
+			name:     "nil url",
+			url:      "",
+			expected: false,
+		},
+		{
+			name:     "huggingface.co main domain",
+			url:      "https://huggingface.co/some/model",
+			expected: true,
+		},
+		{
+			name:     "cdn-lfs.huggingface.co subdomain",
+			url:      "https://cdn-lfs.huggingface.co/repos/abc/123",
+			expected: true,
+		},
+		{
+			name:     "cdn-lfs3.hf.co CDN domain",
+			url:      "https://cdn-lfs3.hf.co/repos/abc/123",
+			expected: true,
+		},
+		{
+			name:     "hf.co shortlink domain",
+			url:      "https://hf.co/model",
+			expected: true,
+		},
+		{
+			name:     "uppercase HuggingFace domain",
+			url:      "https://HUGGINGFACE.CO/model",
+			expected: true,
+		},
+		{
+			name:     "mixed case HF domain",
+			url:      "https://Cdn-Lfs.HF.Co/repos",
+			expected: true,
+		},
+		{
+			name:     "ollama registry",
+			url:      "https://registry.ollama.ai/v2/library/llama3",
+			expected: false,
+		},
+		{
+			name:     "github.com",
+			url:      "https://github.com/ollama/ollama",
+			expected: false,
+		},
+		{
+			name:     "fake huggingface domain",
+			url:      "https://nothuggingface.co/model",
+			expected: false,
+		},
+		{
+			name:     "fake hf domain",
+			url:      "https://nothf.co/model",
+			expected: false,
+		},
+		{
+			name:     "huggingface in path not host",
+			url:      "https://example.com/huggingface.co/model",
+			expected: false,
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			var u *url.URL
+			if tc.url != "" {
+				var err error
+				u, err = url.Parse(tc.url)
+				if err != nil {
+					t.Fatalf("failed to parse URL: %v", err)
+				}
+			}
+			got := isHuggingFaceURL(u)
+			assert.Equal(t, tc.expected, got)
+		})
+	}
+}
+
+func TestGetNumDownloadParts(t *testing.T) {
+	tests := []struct {
+		name        string
+		url         string
+		envValue    string
+		expected    int
+		description string
+	}{
+		{
+			name:        "nil url returns default",
+			url:         "",
+			envValue:    "",
+			expected:    numDownloadParts,
+			description: "nil URL should return standard concurrency",
+		},
+		{
+			name:        "ollama registry returns default",
+			url:         "https://registry.ollama.ai/v2/library/llama3",
+			envValue:    "",
+			expected:    numDownloadParts,
+			description: "Ollama registry should use standard concurrency",
+		},
+		{
+			name:        "huggingface returns reduced default",
+			url:         "https://huggingface.co/model/repo",
+			envValue:    "",
+			expected:    numHFDownloadParts,
+			description: "HuggingFace should use reduced concurrency",
+		},
+		{
+			name:        "hf.co CDN returns reduced default",
+			url:         "https://cdn-lfs3.hf.co/repos/abc/123",
+			envValue:    "",
+			expected:    numHFDownloadParts,
+			description: "HuggingFace CDN should use reduced concurrency",
+		},
+		{
+			name:        "huggingface with env override",
+			url:         "https://huggingface.co/model/repo",
+			envValue:    "2",
+			expected:    2,
+			description: "OLLAMA_HF_CONCURRENCY should override default",
+		},
+		{
+			name:        "huggingface with higher env override",
+			url:         "https://huggingface.co/model/repo",
+			envValue:    "8",
+			expected:    8,
+			description: "OLLAMA_HF_CONCURRENCY can be set higher than default",
+		},
+		{
+			name:        "huggingface with invalid env (non-numeric)",
+			url:         "https://huggingface.co/model/repo",
+			envValue:    "invalid",
+			expected:    numHFDownloadParts,
+			description: "Invalid OLLAMA_HF_CONCURRENCY should fall back to default",
+		},
+		{
+			name:        "huggingface with invalid env (zero)",
+			url:         "https://huggingface.co/model/repo",
+			envValue:    "0",
+			expected:    numHFDownloadParts,
+			description: "Zero OLLAMA_HF_CONCURRENCY should fall back to default",
+		},
+		{
+			name:        "huggingface with invalid env (negative)",
+			url:         "https://huggingface.co/model/repo",
+			envValue:    "-1",
+			expected:    numHFDownloadParts,
+			description: "Negative OLLAMA_HF_CONCURRENCY should fall back to default",
+		},
+		{
+			name:        "non-huggingface ignores env",
+			url:         "https://registry.ollama.ai/v2/library/llama3",
+			envValue:    "2",
+			expected:    numDownloadParts,
+			description: "OLLAMA_HF_CONCURRENCY should not affect non-HF URLs",
+		},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			// Set or clear the environment variable
+			if tc.envValue != "" {
+				t.Setenv("OLLAMA_HF_CONCURRENCY", tc.envValue)
+			}
+
+			var u *url.URL
+			if tc.url != "" {
+				var err error
+				u, err = url.Parse(tc.url)
+				if err != nil {
+					t.Fatalf("failed to parse URL: %v", err)
+				}
+			}
+
+			got := getNumDownloadParts(u)
+			assert.Equal(t, tc.expected, got, tc.description)
+		})
+	}
+}

x/README.md

@@ -1,24 +1,50 @@
-# Experimental Features 
+# Experimental Features
 
 ## MLX Backend
 
 We're working on a new experimental backend based on the [MLX project](https://github.com/ml-explore/mlx)
 
-Support is currently limited to MacOS and Linux with CUDA GPUs.  We're looking to add support for Windows CUDA soon, and other GPU vendors.  To build:
+Support is currently limited to MacOS and Linux with CUDA GPUs. We're looking to add support for Windows CUDA soon, and other GPU vendors.
 
-```
+### Building ollama-mlx
+
+The `ollama-mlx` binary is a separate build of Ollama with MLX support enabled. This enables experimental features like image generation.
+
+#### macOS (Apple Silicon and Intel)
+
+```bash
+# Build MLX backend libraries
 cmake --preset MLX
 cmake --build --preset MLX --parallel
 cmake --install build --component MLX
-go build -tags mlx .
+
+# Build ollama-mlx binary
+go build -tags mlx -o ollama-mlx .
 ```
 
-On linux, use the preset "MLX CUDA 13" or "MLX CUDA 12" to enable CUDA with the default Ollama NVIDIA GPU architectures enabled. 
+#### Linux (CUDA)
+
+On Linux, use the preset "MLX CUDA 13" or "MLX CUDA 12" to enable CUDA with the default Ollama NVIDIA GPU architectures enabled:
+
+```bash
+# Build MLX backend libraries with CUDA support
+cmake --preset 'MLX CUDA 13'
+cmake --build --preset 'MLX CUDA 13' --parallel
+cmake --install build --component MLX
+
+# Build ollama-mlx binary
+CGO_CFLAGS="-O3 -I$(pwd)/build/_deps/mlx-c-src" \
+CGO_LDFLAGS="-L$(pwd)/build/lib/ollama -lmlxc -lmlx" \
+go build -tags mlx -o ollama-mlx .
+```
+
+#### Using build scripts
+
+The build scripts automatically create the `ollama-mlx` binary:
+
+- **macOS**: `./scripts/build_darwin.sh` produces `dist/darwin/ollama-mlx`
+- **Linux**: `./scripts/build_linux.sh` produces `ollama-mlx` in the output archives
 
 ## Image Generation
 
-Based on the experimental MLX backend, we're working on adding imagegen support.  After running the cmake commands above:
-
-```
-go build -o imagegen ./x/imagegen/cmd/engine
-```
+Image generation is built into the `ollama-mlx` binary. Run `ollama-mlx serve` to start the server with image generation support enabled.

x/imagegen/cli.go

@@ -123,11 +123,6 @@ func RegisterFlags(cmd *cobra.Command) {
 // Returns true if it handled the request, false if the caller should continue with normal flow.
 // Supports flags: --width, --height, --steps, --seed, --negative
 func RunCLI(cmd *cobra.Command, name string, prompt string, interactive bool, keepAlive *api.Duration) error {
-	// Verify it's a valid image gen model
-	if ResolveModelName(name) == "" {
-		return fmt.Errorf("unknown image generation model: %s", name)
-	}
-
 	// Get options from flags (with env var defaults)
 	opts := DefaultOptions()
 	if cmd != nil && cmd.Flags() != nil {
@@ -511,10 +506,7 @@ func displayImageInTerminal(imagePath string) bool {
 		// Send in chunks for large images
 		const chunkSize = 4096
 		for i := 0; i < len(encoded); i += chunkSize {
-			end := i + chunkSize
-			if end > len(encoded) {
-				end = len(encoded)
-			}
+			end := min(i+chunkSize, len(encoded))
 			chunk := encoded[i:end]
 
 			if i == 0 {

x/imagegen/server.go

@@ -14,7 +14,9 @@ import (
 	"os"
 	"os/exec"
 	"path/filepath"
+	"runtime"
 	"strconv"
+	"strings"
 	"sync"
 	"time"
 
@@ -70,7 +72,7 @@ func NewServer(modelName string) (*Server, error) {
 		port = rand.Intn(65535-49152) + 49152
 	}
 
-	// Get the ollama executable path
+	// Get the ollama-mlx executable path (in same directory as current executable)
 	exe, err := os.Executable()
 	if err != nil {
 		return nil, fmt.Errorf("unable to lookup executable path: %w", err)
@@ -78,11 +80,42 @@ func NewServer(modelName string) (*Server, error) {
 	if eval, err := filepath.EvalSymlinks(exe); err == nil {
 		exe = eval
 	}
+	mlxExe := filepath.Join(filepath.Dir(exe), "ollama-mlx")
 
-	// Spawn subprocess: ollama runner --image-engine --model <path> --port <port>
-	cmd := exec.Command(exe, "runner", "--image-engine", "--model", modelName, "--port", strconv.Itoa(port))
+	// Spawn subprocess: ollama-mlx runner --image-engine --model <path> --port <port>
+	cmd := exec.Command(mlxExe, "runner", "--image-engine", "--model", modelName, "--port", strconv.Itoa(port))
 	cmd.Env = os.Environ()
 
+	// On Linux, set LD_LIBRARY_PATH to include MLX library directories
+	if runtime.GOOS == "linux" {
+		// Build library paths: start with LibOllamaPath, then add any mlx_* subdirectories
+		libraryPaths := []string{ml.LibOllamaPath}
+		if mlxDirs, err := filepath.Glob(filepath.Join(ml.LibOllamaPath, "mlx_*")); err == nil {
+			libraryPaths = append(libraryPaths, mlxDirs...)
+		}
+
+		// Append existing LD_LIBRARY_PATH if set
+		if existingPath, ok := os.LookupEnv("LD_LIBRARY_PATH"); ok {
+			libraryPaths = append(libraryPaths, filepath.SplitList(existingPath)...)
+		}
+
+		pathEnvVal := strings.Join(libraryPaths, string(filepath.ListSeparator))
+
+		// Update or add LD_LIBRARY_PATH in cmd.Env
+		found := false
+		for i := range cmd.Env {
+			if strings.HasPrefix(cmd.Env[i], "LD_LIBRARY_PATH=") {
+				cmd.Env[i] = "LD_LIBRARY_PATH=" + pathEnvVal
+				found = true
+				break
+			}
+		}
+		if !found {
+			cmd.Env = append(cmd.Env, "LD_LIBRARY_PATH="+pathEnvVal)
+		}
+		slog.Debug("mlx subprocess library path", "LD_LIBRARY_PATH", pathEnvVal)
+	}
+
 	s := &Server{
 		cmd:       cmd,
 		port:      port,
@@ -113,7 +146,7 @@ func NewServer(modelName string) (*Server, error) {
 		}
 	}()
 
-	slog.Info("starting image runner subprocess", "model", modelName, "port", port)
+	slog.Info("starting ollama-mlx image runner subprocess", "exe", mlxExe, "model", modelName, "port", port)
 	if err := cmd.Start(); err != nil {
 		return nil, fmt.Errorf("failed to start image runner: %w", err)
 	}