Add support for IQ1_S, IQ3_S, IQ2_S, IQ4_XS. IQ4_NL

Co-authored-by: ManniX-ITA <20623405+mann1x@users.noreply.github.com>

.gitignore

@@ -12,3 +12,4 @@ ggml-metal.metal
 test_data
 *.crt
 llm/build
+__debug_bin*

README.md

@@ -1,5 +1,5 @@
 <div align="center">
-  <img alt="ollama" height="200px" src="https://github.com/ollama/ollama/assets/3325447/0d0b44e2-8f4a-4e99-9b52-a5c1c741c8f7">
+ <img alt="ollama" height="200px" src="https://github.com/ollama/ollama/assets/3325447/0d0b44e2-8f4a-4e99-9b52-a5c1c741c8f7">
 </div>
 
 # Ollama
@@ -51,7 +51,7 @@ Here are some example models that can be downloaded:
 | ------------------ | ---------- | ----- | ------------------------------ |
 | Llama 3            | 8B         | 4.7GB | `ollama run llama3`            |
 | Llama 3            | 70B        | 40GB  | `ollama run llama3:70b`        |
-| Phi-3              | 3,8B       | 2.3GB | `ollama run phi3`              |
+| Phi-3              | 3.8B       | 2.3GB | `ollama run phi3`              |
 | Mistral            | 7B         | 4.1GB | `ollama run mistral`           |
 | Neural Chat        | 7B         | 4.1GB | `ollama run neural-chat`       |
 | Starling           | 7B         | 4.1GB | `ollama run starling-lm`       |
@@ -173,7 +173,7 @@ I'm a basic program that prints the famous "Hello, world!" message to the consol
 The image features a yellow smiley face, which is likely the central focus of the picture.
 ```
 
-### Pass in prompt as arguments
+### Pass the prompt as an argument
 
 ```
 $ ollama run llama3 "Summarize this file: $(cat README.md)"
@@ -294,7 +294,7 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [RAGFlow: Open-source Retrieval-Augmented Generation engine based on deep document understanding](https://github.com/infiniflow/ragflow)
 - [chat: chat web app for teams](https://github.com/swuecho/chat)
 - [Lobe Chat](https://github.com/lobehub/lobe-chat) with [Integrating Doc](https://lobehub.com/docs/self-hosting/examples/ollama)
-- [Ollama RAG Chatbot: Local Chat with multiples PDFs using Ollama and RAG.](https://github.com/datvodinh/rag-chatbot.git)
+- [Ollama RAG Chatbot: Local Chat with multiple PDFs using Ollama and RAG.](https://github.com/datvodinh/rag-chatbot.git)
 
 ### Terminal
 
@@ -384,4 +384,4 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [Discord-Ollama Chat Bot](https://github.com/kevinthedang/discord-ollama) (Generalized TypeScript Discord Bot w/ Tuning Documentation)
 
 ### Supported backends 
-- [llama.cpp](https://github.com/ggerganov/llama.cpp) project founded by Georgi Gerganov. 
+- [llama.cpp](https://github.com/ggerganov/llama.cpp) project founded by Georgi Gerganov. 

api/client.go

@@ -18,6 +18,7 @@ import (
 	"net/url"
 	"os"
 	"runtime"
+	"strconv"
 	"strings"
 
 	"github.com/ollama/ollama/format"
@@ -57,12 +58,36 @@ func checkError(resp *http.Response, body []byte) error {
 // If the variable is not specified, a default ollama host and port will be
 // used.
 func ClientFromEnvironment() (*Client, error) {
+	ollamaHost, err := GetOllamaHost()
+	if err != nil {
+		return nil, err
+	}
+
+	return &Client{
+		base: &url.URL{
+			Scheme: ollamaHost.Scheme,
+			Host:   net.JoinHostPort(ollamaHost.Host, ollamaHost.Port),
+		},
+		http: http.DefaultClient,
+	}, nil
+}
+
+type OllamaHost struct {
+	Scheme string
+	Host   string
+	Port   string
+}
+
+func GetOllamaHost() (OllamaHost, error) {
 	defaultPort := "11434"
 
-	scheme, hostport, ok := strings.Cut(os.Getenv("OLLAMA_HOST"), "://")
+	hostVar := os.Getenv("OLLAMA_HOST")
+	hostVar = strings.TrimSpace(strings.Trim(strings.TrimSpace(hostVar), "\"'"))
+
+	scheme, hostport, ok := strings.Cut(hostVar, "://")
 	switch {
 	case !ok:
-		scheme, hostport = "http", os.Getenv("OLLAMA_HOST")
+		scheme, hostport = "http", hostVar
 	case scheme == "http":
 		defaultPort = "80"
 	case scheme == "https":
@@ -82,12 +107,14 @@ func ClientFromEnvironment() (*Client, error) {
 		}
 	}
 
-	return &Client{
-		base: &url.URL{
-			Scheme: scheme,
-			Host:   net.JoinHostPort(host, port),
-		},
-		http: http.DefaultClient,
+	if portNum, err := strconv.ParseInt(port, 10, 32); err != nil || portNum > 65535 || portNum < 0 {
+		return OllamaHost{}, ErrInvalidHostPort
+	}
+
+	return OllamaHost{
+		Scheme: scheme,
+		Host:   host,
+		Port:   port,
 	}, nil
 }
 

api/client_test.go

@@ -1,6 +1,12 @@
 package api
 
-import "testing"
+import (
+	"fmt"
+	"net"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
 
 func TestClientFromEnvironment(t *testing.T) {
 	type testCase struct {
@@ -40,4 +46,40 @@ func TestClientFromEnvironment(t *testing.T) {
 			}
 		})
 	}
+
+	hostTestCases := map[string]*testCase{
+		"empty":               {value: "", expect: "127.0.0.1:11434"},
+		"only address":        {value: "1.2.3.4", expect: "1.2.3.4:11434"},
+		"only port":           {value: ":1234", expect: ":1234"},
+		"address and port":    {value: "1.2.3.4:1234", expect: "1.2.3.4:1234"},
+		"hostname":            {value: "example.com", expect: "example.com:11434"},
+		"hostname and port":   {value: "example.com:1234", expect: "example.com:1234"},
+		"zero port":           {value: ":0", expect: ":0"},
+		"too large port":      {value: ":66000", err: ErrInvalidHostPort},
+		"too small port":      {value: ":-1", err: ErrInvalidHostPort},
+		"ipv6 localhost":      {value: "[::1]", expect: "[::1]:11434"},
+		"ipv6 world open":     {value: "[::]", expect: "[::]:11434"},
+		"ipv6 no brackets":    {value: "::1", expect: "[::1]:11434"},
+		"ipv6 + port":         {value: "[::1]:1337", expect: "[::1]:1337"},
+		"extra space":         {value: " 1.2.3.4 ", expect: "1.2.3.4:11434"},
+		"extra quotes":        {value: "\"1.2.3.4\"", expect: "1.2.3.4:11434"},
+		"extra space+quotes":  {value: " \" 1.2.3.4 \" ", expect: "1.2.3.4:11434"},
+		"extra single quotes": {value: "'1.2.3.4'", expect: "1.2.3.4:11434"},
+	}
+
+	for k, v := range hostTestCases {
+		t.Run(k, func(t *testing.T) {
+			t.Setenv("OLLAMA_HOST", v.value)
+
+			oh, err := GetOllamaHost()
+			if err != v.err {
+				t.Fatalf("expected %s, got %s", v.err, err)
+			}
+
+			if err == nil {
+				host := net.JoinHostPort(oh.Host, oh.Port)
+				assert.Equal(t, v.expect, host, fmt.Sprintf("%s: expected %s, got %s", k, v.expect, host))
+			}
+		})
+	}
 }

api/types.go

@@ -309,6 +309,7 @@ func (m *Metrics) Summary() {
 }
 
 var ErrInvalidOpts = errors.New("invalid options")
+var ErrInvalidHostPort = errors.New("invalid port specified in OLLAMA_HOST")
 
 func (opts *Options) FromMap(m map[string]interface{}) error {
 	valueOpts := reflect.ValueOf(opts).Elem() // names of the fields in the options struct

app/.gitignore

@@ -1,2 +1 @@
 ollama.syso
-app

app/AppDelegate.h

@@ -1,7 +0,0 @@
-#import <Cocoa/Cocoa.h>
-
-@interface AppDelegate : NSObject <NSApplicationDelegate>
-
-- (void)applicationDidFinishLaunching:(NSNotification *)aNotification;
-
-@end

app/README.md

@@ -1,6 +1,10 @@
 # Ollama App
 
-## macOS
+## Linux
+
+TODO
+
+## MacOS
 
 TODO
 

app/app_darwin.go

@@ -1,76 +0,0 @@
-package main
-
-// #cgo CFLAGS: -x objective-c
-// #cgo LDFLAGS: -framework Cocoa -framework LocalAuthentication -framework ServiceManagement
-// #include "app_darwin.h"
-import "C"
-import (
-	"context"
-	"fmt"
-	"log/slog"
-	"os"
-	"path/filepath"
-	"syscall"
-)
-
-func init() {
-	home, err := os.UserHomeDir()
-	if err != nil {
-		panic(err)
-	}
-
-	ServerLogFile = filepath.Join(home, ".ollama", "logs", "server.log")
-}
-
-func run() {
-	initLogging()
-	slog.Info("ollama macOS app started")
-
-	// Ask to move to applications directory
-	moving := C.askToMoveToApplications()
-	if moving {
-		return
-	}
-
-	C.killOtherInstances()
-
-	code := C.installSymlink()
-	if code != 0 {
-		slog.Error("Failed to install symlink")
-	}
-
-	exe, err := os.Executable()
-	if err != nil {
-		panic(err)
-	}
-
-	var options ServerOptions
-
-	ctx, cancel := context.WithCancel(context.Background())
-	var done chan int
-
-	done, err = SpawnServer(ctx, filepath.Join(filepath.Dir(exe), "..", "Resources", "ollama"), options)
-	if err != nil {
-		slog.Error(fmt.Sprintf("Failed to spawn ollama server %s", err))
-		done = make(chan int, 1)
-		done <- 1
-	}
-
-	// Run the native macOS app
-	// Note: this will block until the app is closed
-	C.run()
-
-	slog.Info("ollama macOS app closed")
-
-	cancel()
-	slog.Info("Waiting for ollama server to shutdown...")
-	if done != nil {
-		<-done
-	}
-	slog.Info("Ollama app exiting")
-}
-
-//export Quit
-func Quit() {
-	syscall.Kill(os.Getpid(), syscall.SIGTERM)
-}

app/app_darwin.h

@@ -1,13 +0,0 @@
-#import <Cocoa/Cocoa.h>
-
-@interface AppDelegate : NSObject <NSApplicationDelegate>
-- (void)applicationDidFinishLaunching:(NSNotification *)aNotification;
-@end
-
-void run();
-void killOtherInstances();
-bool askToMoveToApplications();
-int createSymlinkWithAuthorization();
-int installSymlink();
-extern void Restart();
-extern void Quit();

app/app_darwin.m

@@ -1,282 +0,0 @@
-#import <AppKit/AppKit.h>
-#import <Cocoa/Cocoa.h>
-#import <CoreServices/CoreServices.h>
-#import <Security/Security.h>
-#import <ServiceManagement/ServiceManagement.h>
-#import "app_darwin.h"
-
-@interface AppDelegate ()
-
-@property (strong, nonatomic) NSStatusItem *statusItem;
-
-@end
-
-@implementation AppDelegate
-
-- (void)applicationDidFinishLaunching:(NSNotification *)aNotification {
-    // show status menu
-    NSMenu *menu = [[NSMenu alloc] init];
-
-    NSMenuItem *aboutMenuItem = [[NSMenuItem alloc] initWithTitle:@"About Ollama" action:@selector(aboutOllama) keyEquivalent:@""];
-    [aboutMenuItem setTarget:self];
-    [menu addItem:aboutMenuItem];
-
-    // Settings submenu
-    NSMenu *settingsMenu = [[NSMenu alloc] initWithTitle:@"Settings"];
-
-    // Submenu items
-    NSMenuItem *chooseModelDirectoryItem = [[NSMenuItem alloc] initWithTitle:@"Choose model directory..." action:@selector(chooseModelDirectory) keyEquivalent:@""];
-    [chooseModelDirectoryItem setTarget:self];
-    [chooseModelDirectoryItem setEnabled:YES];
-    [settingsMenu addItem:chooseModelDirectoryItem];
-
-    NSMenuItem *exposeExternallyItem = [[NSMenuItem alloc] initWithTitle:@"Allow external connections" action:@selector(toggleExposeExternally:) keyEquivalent:@""];
-    [exposeExternallyItem setTarget:self];
-    [exposeExternallyItem setState:NSOffState]; // Set initial state to off
-    [exposeExternallyItem setEnabled:YES];
-    [settingsMenu addItem:exposeExternallyItem];
-
-    NSMenuItem *allowCrossOriginItem = [[NSMenuItem alloc] initWithTitle:@"Allow browser requests" action:@selector(toggleCrossOrigin:) keyEquivalent:@""];
-    [allowCrossOriginItem setTarget:self];
-    [allowCrossOriginItem setState:NSOffState]; // Set initial state to off
-    [allowCrossOriginItem setEnabled:YES];
-    [settingsMenu addItem:allowCrossOriginItem];
-
-    NSMenuItem *settingsMenuItem = [[NSMenuItem alloc] initWithTitle:@"Settings" action:nil keyEquivalent:@""];
-    [settingsMenuItem setSubmenu:settingsMenu];
-    [menu addItem:settingsMenuItem];
-
-    [menu addItemWithTitle:@"Quit Ollama" action:@selector(quit) keyEquivalent:@"q"];
-
-    self.statusItem = [[NSStatusBar systemStatusBar] statusItemWithLength:NSVariableStatusItemLength];
-    [self.statusItem addObserver:self forKeyPath:@"button.effectiveAppearance" options:NSKeyValueObservingOptionNew|NSKeyValueObservingOptionInitial context:nil];
-
-    self.statusItem.menu = menu;
-    [self showIcon];
-}
-
-- (void)aboutOllama {
-    [[NSApplication sharedApplication] orderFrontStandardAboutPanel:nil];
-}
-
-- (void)toggleCrossOrigin:(id)sender {
-    NSMenuItem *item = (NSMenuItem *)sender;
-    if ([item state] == NSOffState) {
-        // Do something when cross-origin requests are allowed
-        [item setState:NSOnState];
-    } else {
-        // Do something when cross-origin requests are disallowed
-        [item setState:NSOffState];
-    }
-}
-
-- (void)toggleExposeExternally:(id)sender {
-    NSMenuItem *item = (NSMenuItem *)sender;
-    if ([item state] == NSOffState) {
-        // Do something when Ollama is exposed externally
-        [item setState:NSOnState];
-    } else {
-        // Do something when Ollama is not exposed externally
-        [item setState:NSOffState];
-    }
-}
-
-- (void)chooseModelDirectory {
-    NSOpenPanel *openPanel = [NSOpenPanel openPanel];
-    [openPanel setCanChooseFiles:NO];
-    [openPanel setCanChooseDirectories:YES];
-    [openPanel setAllowsMultipleSelection:NO];
-
-    NSInteger result = [openPanel runModal];
-    if (result == NSModalResponseOK) {
-        NSURL *selectedDirectoryURL = [openPanel URLs].firstObject;
-        // Do something with the selected directory URL
-    }
-}
-
--(void) showIcon {
-    NSAppearance* appearance = self.statusItem.button.effectiveAppearance;
-    NSString* appearanceName = (NSString*)(appearance.name);
-    NSString* iconName = [[appearanceName lowercaseString] containsString:@"dark"] ? @"iconDark" : @"icon";
-    NSImage* statusImage = [NSImage imageNamed:iconName];
-    [statusImage setTemplate:YES];
-    self.statusItem.button.image = statusImage;
-}
-
--(void)observeValueForKeyPath:(NSString *)keyPath ofObject:(id)object change:(NSDictionary<NSKeyValueChangeKey,id> *)change context:(void *)context {
-    [self showIcon];
-}
-
-- (void)quit {
-    [NSApp stop:nil];
-}
-
-@end
-
-void run() {
-    @autoreleasepool {
-        [NSApplication sharedApplication];
-        AppDelegate *appDelegate = [[AppDelegate alloc] init];
-        [NSApp setDelegate:appDelegate];
-        [NSApp run];
-    }
-}
-
-// killOtherInstances kills all other instances of the app currently
-// running. This way we can ensure that only the most recently started
-// instance of Ollama is running
-void killOtherInstances() {
-    pid_t pid = getpid();
-    NSArray *all = [[NSWorkspace sharedWorkspace] runningApplications];
-    NSMutableArray *apps = [NSMutableArray array];
-
-    for (NSRunningApplication *app in all) {
-        if ([app.bundleIdentifier isEqualToString:[[NSBundle mainBundle] bundleIdentifier]] ||
-            [app.bundleIdentifier isEqualToString:@"ai.ollama.ollama"] ||
-            [app.bundleIdentifier isEqualToString:@"com.electron.ollama"]) {
-            if (app.processIdentifier != pid) {
-                [apps addObject:app];
-            }
-        }
-    }
-
-    for (NSRunningApplication *app in apps) {
-        kill(app.processIdentifier, SIGTERM);
-    }
-
-    NSDate *startTime = [NSDate date];
-    for (NSRunningApplication *app in apps) {
-        while (!app.terminated) {
-            if (-[startTime timeIntervalSinceNow] >= 5) {
-                kill(app.processIdentifier, SIGKILL);
-                break;
-            }
-
-            [[NSRunLoop currentRunLoop] runUntilDate:[NSDate dateWithTimeIntervalSinceNow:0.1]];
-        }
-    }
-}
-
-bool askToMoveToApplications() {
-    NSString *bundlePath = [[NSBundle mainBundle] bundlePath];
-    if ([bundlePath hasPrefix:@"/Applications"]) {
-        return false;
-    }
-
-    NSAlert *alert = [[NSAlert alloc] init];
-    [alert setMessageText:@"Move to Applications?"];
-    [alert setInformativeText:@"Ollama works best when run from the Applications directory."];
-    [alert addButtonWithTitle:@"Move to Applications"];
-    [alert addButtonWithTitle:@"Don't move"];
-
-    [NSApp activateIgnoringOtherApps:YES];
-
-    if ([alert runModal] != NSAlertFirstButtonReturn) {
-        return false;
-    }
-
-    // move to applications
-    NSString *applicationsPath = @"/Applications";
-    NSString *newPath = [applicationsPath stringByAppendingPathComponent:@"Ollama.app"];
-    NSFileManager *fileManager = [NSFileManager defaultManager];
-
-    // Check if the newPath already exists
-    if ([fileManager fileExistsAtPath:newPath]) {
-        NSError *removeError = nil;
-        [fileManager removeItemAtPath:newPath error:&removeError];
-        if (removeError) {
-            NSLog(@"Error removing file at %@: %@", newPath, removeError);
-            return false; // or handle the error
-        }
-    }
-
-    NSError *moveError = nil;
-    [fileManager moveItemAtPath:bundlePath toPath:newPath error:&moveError];
-    if (moveError) {
-        NSLog(@"Error moving file from %@ to %@: %@", bundlePath, newPath, moveError);
-        return false;
-    }
-
-    NSLog(@"Opening %@", newPath);
-    NSError *error = nil;
-    NSWorkspace *workspace = [NSWorkspace sharedWorkspace];
-#pragma clang diagnostic ignored "-Wdeprecated-declarations"
-    [workspace launchApplicationAtURL:[NSURL fileURLWithPath:newPath]
-               options:NSWorkspaceLaunchNewInstance | NSWorkspaceLaunchDefault
-               configuration:@{}
-               error:&error];
-
-    return true;
-}
-
-int installSymlink() {
-    NSString *linkPath = @"/usr/local/bin/ollama";
-    NSError *error = nil;
-
-    NSFileManager *fileManager = [NSFileManager defaultManager];
-    NSString *symlinkPath = [fileManager destinationOfSymbolicLinkAtPath:linkPath error:&error];
-    NSString *bundlePath = [[NSBundle mainBundle] bundlePath];
-    NSString *execPath = [[NSBundle mainBundle] executablePath];
-    NSString *resPath = [[NSBundle mainBundle] pathForResource:@"ollama" ofType:nil];
-
-    // if the symlink already exists and points to the right place, don't prompt
-    if ([symlinkPath isEqualToString:resPath]) {
-        NSLog(@"symbolic link already exists and points to the right place");
-        return 0;
-    }
-
-    NSString *authorizationPrompt = @"Ollama is trying to install its command line interface (CLI) tool.";
-
-    AuthorizationRef auth = NULL;
-    OSStatus createStatus = AuthorizationCreate(NULL, kAuthorizationEmptyEnvironment, kAuthorizationFlagDefaults, &auth);
-    if (createStatus != errAuthorizationSuccess) {
-        NSLog(@"Error creating authorization");
-        return -1;
-    }
-
-    NSString * bundleIdentifier = [[NSBundle mainBundle] bundleIdentifier];
-    NSString *rightNameString = [NSString stringWithFormat:@"%@.%@", bundleIdentifier, @"auth3"];
-    const char *rightName = rightNameString.UTF8String;
-
-    OSStatus getRightResult = AuthorizationRightGet(rightName, NULL);
-    if (getRightResult == errAuthorizationDenied) {
-        if (AuthorizationRightSet(auth, rightName, (__bridge CFTypeRef _Nonnull)(@(kAuthorizationRuleAuthenticateAsAdmin)), (__bridge CFStringRef _Nullable)(authorizationPrompt), NULL, NULL) != errAuthorizationSuccess) {
-            NSLog(@"Failed to set right");
-            return -1;
-        }
-    }
-
-    AuthorizationItem right = { .name = rightName, .valueLength = 0, .value = NULL, .flags = 0 };
-    AuthorizationRights rights = { .count = 1, .items = &right };
-    AuthorizationFlags flags = (AuthorizationFlags)(kAuthorizationFlagExtendRights | kAuthorizationFlagInteractionAllowed);
-    AuthorizationItem iconAuthorizationItem = {.name = kAuthorizationEnvironmentIcon, .valueLength = 0, .value = NULL, .flags = 0};
-    AuthorizationEnvironment authorizationEnvironment = {.count = 0, .items = NULL};
-
-    BOOL failedToUseSystemDomain = NO;
-    OSStatus copyStatus = AuthorizationCopyRights(auth, &rights, &authorizationEnvironment, flags, NULL);
-    if (copyStatus != errAuthorizationSuccess) {
-        failedToUseSystemDomain = YES;
-
-        if (copyStatus == errAuthorizationCanceled) {
-            NSLog(@"User cancelled authorization");
-            return -1;
-        } else {
-            NSLog(@"Failed copying system domain rights: %d", copyStatus);
-            return -1;
-        }
-    }
-
-    const char *toolPath = "/bin/ln";
-    const char *args[] = {"-s", "-F", [resPath UTF8String], "/usr/local/bin/ollama", NULL};
-    FILE *pipe = NULL;
-
-#pragma clang diagnostic ignored "-Wdeprecated-declarations"
-    OSStatus status = AuthorizationExecuteWithPrivileges(auth, toolPath, kAuthorizationFlagDefaults, (char *const *)args, &pipe);
-    if (status != errAuthorizationSuccess) {
-        NSLog(@"Failed to create symlink");
-        return -1;
-    }
-
-    AuthorizationFree(auth, kAuthorizationFlagDestroyRights);
-    return 0;
-}

app/app_windows.go

@@ -1,166 +0,0 @@
-package main
-
-import (
-	"context"
-	"errors"
-	"fmt"
-	"log"
-	"log/slog"
-	"os"
-	"os/exec"
-	"os/signal"
-	"path/filepath"
-	"strings"
-	"syscall"
-
-	"github.com/ollama/ollama/app/lifecycle"
-	"github.com/ollama/ollama/app/store"
-	"github.com/ollama/ollama/app/tray"
-	"github.com/ollama/ollama/app/updater"
-)
-
-func init() {
-	AppName += ".exe"
-	CLIName += ".exe"
-	// Logs, configs, downloads go to LOCALAPPDATA
-	localAppData := os.Getenv("LOCALAPPDATA")
-	AppDataDir = filepath.Join(localAppData, "Ollama")
-	AppLogFile = filepath.Join(AppDataDir, "app.log")
-	ServerLogFile = filepath.Join(AppDataDir, "server.log")
-
-	// Executables are stored in APPDATA
-	AppDir = filepath.Join(localAppData, "Programs", "Ollama")
-
-	// Make sure we have PATH set correctly for any spawned children
-	paths := strings.Split(os.Getenv("PATH"), ";")
-	// Start with whatever we find in the PATH/LD_LIBRARY_PATH
-	found := false
-	for _, path := range paths {
-		d, err := filepath.Abs(path)
-		if err != nil {
-			continue
-		}
-		if strings.EqualFold(AppDir, d) {
-			found = true
-		}
-	}
-	if !found {
-		paths = append(paths, AppDir)
-
-		pathVal := strings.Join(paths, ";")
-		slog.Debug("setting PATH=" + pathVal)
-		err := os.Setenv("PATH", pathVal)
-		if err != nil {
-			slog.Error(fmt.Sprintf("failed to update PATH: %s", err))
-		}
-	}
-
-	// Make sure our logging dir exists
-	_, err := os.Stat(AppDataDir)
-	if errors.Is(err, os.ErrNotExist) {
-		if err := os.MkdirAll(AppDataDir, 0o755); err != nil {
-			slog.Error(fmt.Sprintf("create ollama dir %s: %v", AppDataDir, err))
-		}
-	}
-}
-
-func ShowLogs() {
-	cmd_path := "c:\\Windows\\system32\\cmd.exe"
-	slog.Debug(fmt.Sprintf("viewing logs with start %s", AppDataDir))
-	cmd := exec.Command(cmd_path, "/c", "start", AppDataDir)
-	cmd.SysProcAttr = &syscall.SysProcAttr{HideWindow: false, CreationFlags: 0x08000000}
-	err := cmd.Start()
-	if err != nil {
-		slog.Error(fmt.Sprintf("Failed to open log dir: %s", err))
-	}
-}
-
-func Start() {
-	cmd_path := "c:\\Windows\\system32\\cmd.exe"
-	slog.Debug(fmt.Sprintf("viewing logs with start %s", AppDataDir))
-	cmd := exec.Command(cmd_path, "/c", "start", AppDataDir)
-	cmd.SysProcAttr = &syscall.SysProcAttr{HideWindow: false, CreationFlags: 0x08000000}
-	err := cmd.Start()
-	if err != nil {
-		slog.Error(fmt.Sprintf("Failed to open log dir: %s", err))
-	}
-}
-
-func run() {
-	initLogging()
-
-	slog.Info("ollama windows app started")
-
-	ctx, cancel := context.WithCancel(context.Background())
-	var done chan int
-
-	t, err := tray.NewTray()
-	if err != nil {
-		log.Fatalf("Failed to start: %s", err)
-	}
-	callbacks := t.GetCallbacks()
-
-	signals := make(chan os.Signal, 1)
-	signal.Notify(signals, syscall.SIGINT, syscall.SIGTERM)
-
-	go func() {
-		slog.Debug("starting callback loop")
-		for {
-			select {
-			case <-callbacks.Quit:
-				slog.Debug("quit called")
-				t.Quit()
-			case <-signals:
-				slog.Debug("shutting down due to signal")
-				t.Quit()
-			case <-callbacks.Update:
-				err := updater.DoUpgrade(cancel, done)
-				if err != nil {
-					slog.Warn(fmt.Sprintf("upgrade attempt failed: %s", err))
-				}
-			case <-callbacks.ShowLogs:
-				ShowLogs()
-			case <-callbacks.DoFirstUse:
-				err := lifecycle.GetStarted()
-				if err != nil {
-					slog.Warn(fmt.Sprintf("Failed to launch getting started shell: %s", err))
-				}
-			}
-		}
-	}()
-
-	if !store.GetFirstTimeRun() {
-		slog.Debug("First time run")
-		err = t.DisplayFirstUseNotification()
-		if err != nil {
-			slog.Debug(fmt.Sprintf("XXX failed to display first use notification %v", err))
-		}
-		store.SetFirstTimeRun(true)
-	} else {
-		slog.Debug("Not first time, skipping first run notification")
-	}
-
-	if isServerRunning(ctx) {
-		slog.Info("Detected another instance of ollama running, exiting")
-		os.Exit(1)
-	}
-
-	done, err = SpawnServer(ctx, CLIName)
-	if err != nil {
-		// TODO - should we retry in a backoff loop?
-		// TODO - should we pop up a warning and maybe add a menu item to view application logs?
-		slog.Error(fmt.Sprintf("Failed to spawn ollama server %s", err))
-		done = make(chan int, 1)
-		done <- 1
-	}
-
-	updater.StartBackgroundUpdaterChecker(ctx, t.UpdateAvailable)
-
-	t.Run()
-	cancel()
-	slog.Info("Waiting for ollama server to shutdown...")
-	if done != nil {
-		<-done
-	}
-	slog.Info("Ollama app exiting")
-}

app/darwin/Ollama.app/Contents/Info.plist

@@ -1,40 +0,0 @@
-<?xml version="1.0" encoding="UTF-8"?>
-<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
-<plist version="1.0">
-  <dict>
-    <key>CFBundleDisplayName</key>
-    <string>Ollama</string>
-    <key>CFBundleExecutable</key>
-    <string>Ollama</string>
-    <key>CFBundleIconFile</key>
-    <string>icon.icns</string>
-    <key>CFBundleIdentifier</key>
-    <string>com.ollama.ollama</string>
-    <key>CFBundleInfoDictionaryVersion</key>
-    <string>6.0</string>
-    <key>CFBundleName</key>
-    <string>Ollama</string>
-    <key>CFBundlePackageType</key>
-    <string>APPL</string>
-    <key>CFBundleShortVersionString</key>
-    <string>0.0.0</string>
-    <key>CFBundleVersion</key>
-    <string>0.0.0</string>
-    <key>DTCompiler</key>
-    <string>com.apple.compilers.llvm.clang.1_0</string>
-    <key>DTSDKBuild</key>
-    <string>22E245</string>
-    <key>DTSDKName</key>
-    <string>macosx13.3</string>
-    <key>DTXcode</key>
-    <string>1431</string>
-    <key>DTXcodeBuild</key>
-    <string>14E300c</string>
-    <key>LSApplicationCategoryType</key>
-    <string>public.app-category.developer-tools</string>
-    <key>LSMinimumSystemVersion</key>
-    <string>11.0</string>
-    <key>LSUIElement</key>
-    <true/>
-  </dict>
-</plist>

app/lifecycle/getstarted_nonwindows.go

@@ -1,3 +1,5 @@
+//go:build !windows
+
 package lifecycle
 
 import "fmt"

app/lifecycle/lifecycle.go

@@ -0,0 +1,92 @@
+package lifecycle
+
+import (
+	"context"
+	"fmt"
+	"log"
+	"log/slog"
+	"os"
+	"os/signal"
+	"syscall"
+
+	"github.com/ollama/ollama/app/store"
+	"github.com/ollama/ollama/app/tray"
+)
+
+func Run() {
+	InitLogging()
+
+	ctx, cancel := context.WithCancel(context.Background())
+	var done chan int
+
+	t, err := tray.NewTray()
+	if err != nil {
+		log.Fatalf("Failed to start: %s", err)
+	}
+	callbacks := t.GetCallbacks()
+
+	signals := make(chan os.Signal, 1)
+	signal.Notify(signals, syscall.SIGINT, syscall.SIGTERM)
+
+	go func() {
+		slog.Debug("starting callback loop")
+		for {
+			select {
+			case <-callbacks.Quit:
+				slog.Debug("quit called")
+				t.Quit()
+			case <-signals:
+				slog.Debug("shutting down due to signal")
+				t.Quit()
+			case <-callbacks.Update:
+				err := DoUpgrade(cancel, done)
+				if err != nil {
+					slog.Warn(fmt.Sprintf("upgrade attempt failed: %s", err))
+				}
+			case <-callbacks.ShowLogs:
+				ShowLogs()
+			case <-callbacks.DoFirstUse:
+				err := GetStarted()
+				if err != nil {
+					slog.Warn(fmt.Sprintf("Failed to launch getting started shell: %s", err))
+				}
+			}
+		}
+	}()
+
+	// Are we first use?
+	if !store.GetFirstTimeRun() {
+		slog.Debug("First time run")
+		err = t.DisplayFirstUseNotification()
+		if err != nil {
+			slog.Debug(fmt.Sprintf("XXX failed to display first use notification %v", err))
+		}
+		store.SetFirstTimeRun(true)
+	} else {
+		slog.Debug("Not first time, skipping first run notification")
+	}
+
+	if IsServerRunning(ctx) {
+		slog.Info("Detected another instance of ollama running, exiting")
+		os.Exit(1)
+	} else {
+		done, err = SpawnServer(ctx, CLIName)
+		if err != nil {
+			// TODO - should we retry in a backoff loop?
+			// TODO - should we pop up a warning and maybe add a menu item to view application logs?
+			slog.Error(fmt.Sprintf("Failed to spawn ollama server %s", err))
+			done = make(chan int, 1)
+			done <- 1
+		}
+	}
+
+	StartBackgroundUpdaterChecker(ctx, t.UpdateAvailable)
+
+	t.Run()
+	cancel()
+	slog.Info("Waiting for ollama server to shutdown...")
+	if done != nil {
+		<-done
+	}
+	slog.Info("Ollama app exiting")
+}

app/lifecycle/logging.go

@@ -1,4 +1,4 @@
-package main
+package lifecycle
 
 import (
 	"fmt"
@@ -7,7 +7,7 @@ import (
 	"path/filepath"
 )
 
-func initLogging() {
+func InitLogging() {
 	level := slog.LevelInfo
 
 	if debug := os.Getenv("OLLAMA_DEBUG"); debug != "" {
@@ -41,4 +41,6 @@ func initLogging() {
 	})
 
 	slog.SetDefault(slog.New(handler))
+
+	slog.Info("ollama app started")
 }

app/lifecycle/logging_nonwindows.go

@@ -0,0 +1,9 @@
+//go:build !windows
+
+package lifecycle
+
+import "log/slog"
+
+func ShowLogs() {
+	slog.Warn("ShowLogs not yet implemented")
+}

app/lifecycle/logging_windows.go

@@ -0,0 +1,19 @@
+package lifecycle
+
+import (
+	"fmt"
+	"log/slog"
+	"os/exec"
+	"syscall"
+)
+
+func ShowLogs() {
+	cmd_path := "c:\\Windows\\system32\\cmd.exe"
+	slog.Debug(fmt.Sprintf("viewing logs with start %s", AppDataDir))
+	cmd := exec.Command(cmd_path, "/c", "start", AppDataDir)
+	cmd.SysProcAttr = &syscall.SysProcAttr{HideWindow: false, CreationFlags: 0x08000000}
+	err := cmd.Start()
+	if err != nil {
+		slog.Error(fmt.Sprintf("Failed to open log dir: %s", err))
+	}
+}

app/lifecycle/paths.go

@@ -70,5 +70,10 @@ func init() {
 			}
 		}
 
+	} else if runtime.GOOS == "darwin" {
+		// TODO
+		AppName += ".app"
+		// } else if runtime.GOOS == "linux" {
+		// TODO
 	}
 }

app/lifecycle/server.go

@@ -1,4 +1,4 @@
-package main
+package lifecycle
 
 import (
 	"context"
@@ -14,28 +14,37 @@ import (
 	"github.com/ollama/ollama/api"
 )
 
-type ServerOptions struct {
-	Cors       bool
-	Expose     bool
-	ModelsPath string
+func getCLIFullPath(command string) string {
+	cmdPath := ""
+	appExe, err := os.Executable()
+	if err == nil {
+		cmdPath = filepath.Join(filepath.Dir(appExe), command)
+		_, err := os.Stat(cmdPath)
+		if err == nil {
+			return cmdPath
+		}
+	}
+	cmdPath, err = exec.LookPath(command)
+	if err == nil {
+		_, err := os.Stat(cmdPath)
+		if err == nil {
+			return cmdPath
+		}
+	}
+	pwd, err := os.Getwd()
+	if err == nil {
+		cmdPath = filepath.Join(pwd, command)
+		_, err = os.Stat(cmdPath)
+		if err == nil {
+			return cmdPath
+		}
+	}
+
+	return command
 }
 
-func start(ctx context.Context, command string, options ServerOptions) (*exec.Cmd, error) {
-	cmd := getCmd(ctx, command)
-
-	// set environment variables
-	if options.ModelsPath != "" {
-		cmd.Env = append(cmd.Env, fmt.Sprintf("OLLAMA_MODELS=%s", options.ModelsPath))
-	}
-
-	if options.Cors {
-		cmd.Env = append(cmd.Env, "OLLAMA_ORIGINS=*")
-	}
-
-	if options.Expose {
-		cmd.Env = append(cmd.Env, "OLLAMA_HOST=0.0.0.0")
-	}
-
+func start(ctx context.Context, command string) (*exec.Cmd, error) {
+	cmd := getCmd(ctx, getCLIFullPath(command))
 	stdout, err := cmd.StdoutPipe()
 	if err != nil {
 		return nil, fmt.Errorf("failed to spawn server stdout pipe: %w", err)
@@ -50,6 +59,20 @@ func start(ctx context.Context, command string, options ServerOptions) (*exec.Cm
 	if err != nil {
 		return nil, fmt.Errorf("failed to create server log: %w", err)
 	}
+
+	logDir := filepath.Dir(ServerLogFile)
+	_, err = os.Stat(logDir)
+	if err != nil {
+		if !errors.Is(err, os.ErrNotExist) {
+			return nil, fmt.Errorf("stat ollama server log dir %s: %v", logDir, err)
+
+		}
+
+		if err := os.MkdirAll(logDir, 0o755); err != nil {
+			return nil, fmt.Errorf("create ollama server log dir %s: %v", logDir, err)
+		}
+	}
+
 	go func() {
 		defer logFile.Close()
 		io.Copy(logFile, stdout) //nolint:errcheck
@@ -103,25 +126,20 @@ func start(ctx context.Context, command string, options ServerOptions) (*exec.Cm
 	return cmd, nil
 }
 
-func SpawnServer(ctx context.Context, command string, options ServerOptions) (chan int, error) {
-	logDir := filepath.Dir(ServerLogFile)
-	_, err := os.Stat(logDir)
-	if errors.Is(err, os.ErrNotExist) {
-		if err := os.MkdirAll(logDir, 0o755); err != nil {
-			return nil, fmt.Errorf("create ollama server log dir %s: %v", logDir, err)
-		}
-	}
-
+func SpawnServer(ctx context.Context, command string) (chan int, error) {
 	done := make(chan int)
 
 	go func() {
 		// Keep the server running unless we're shuttind down the app
 		crashCount := 0
 		for {
-			slog.Info(fmt.Sprintf("starting server..."))
-			cmd, err := start(ctx, command, options)
+			slog.Info("starting server...")
+			cmd, err := start(ctx, command)
 			if err != nil {
+				crashCount++
 				slog.Error(fmt.Sprintf("failed to start server %s", err))
+				time.Sleep(500 * time.Millisecond * time.Duration(crashCount))
+				continue
 			}
 
 			cmd.Wait() //nolint:errcheck
@@ -147,7 +165,7 @@ func SpawnServer(ctx context.Context, command string, options ServerOptions) (ch
 	return done, nil
 }
 
-func isServerRunning(ctx context.Context) bool {
+func IsServerRunning(ctx context.Context) bool {
 	client, err := api.ClientFromEnvironment()
 	if err != nil {
 		slog.Info("unable to connect to server")

app/lifecycle/server_unix.go

@@ -1,4 +1,6 @@
-package main
+//go:build !windows
+
+package lifecycle
 
 import (
 	"context"

app/lifecycle/server_windows.go

@@ -1,4 +1,4 @@
-package main
+package lifecycle
 
 import (
 	"context"

app/lifecycle/updater.go

@@ -1,4 +1,4 @@
-package updater
+package lifecycle
 
 import (
 	"context"
@@ -22,10 +22,6 @@ import (
 	"github.com/ollama/ollama/version"
 )
 
-var (
-	UpdateStageDir string
-)
-
 var (
 	UpdateCheckURLBase  = "https://ollama.com/api/update"
 	UpdateDownloaded    = false
@@ -127,7 +123,7 @@ func DownloadNewRelease(ctx context.Context, updateResp UpdateResponse) error {
 		slog.Debug("no etag detected, falling back to filename based dedup")
 		etag = "_"
 	}
-	filename := "OllamaSetup.exe"
+	filename := Installer
 	_, params, err := mime.ParseMediaType(resp.Header.Get("content-disposition"))
 	if err == nil {
 		filename = params["filename"]

app/lifecycle/updater_nonwindows.go

@@ -1,4 +1,6 @@
-package updater
+//go:build !windows
+
+package lifecycle
 
 import (
 	"context"

app/lifecycle/updater_windows.go

@@ -1,4 +1,4 @@
-package updater
+package lifecycle
 
 import (
 	"context"
@@ -9,13 +9,7 @@ import (
 	"path/filepath"
 )
 
-func init() {
-	UpdateStageDir = filepath.Join(os.Getenv("LOCALAPPDATA"), "Ollama", "updates")
-}
-
 func DoUpgrade(cancel context.CancelFunc, done chan int) error {
-	logFile := filepath.Join(os.Getenv("LOCALAPPDATA"), "Ollama", "upgrade.log")
-
 	files, err := filepath.Glob(filepath.Join(UpdateStageDir, "*", "*.exe")) // TODO generalize for multiplatform
 	if err != nil {
 		return fmt.Errorf("failed to lookup downloads: %s", err)
@@ -29,13 +23,13 @@ func DoUpgrade(cancel context.CancelFunc, done chan int) error {
 	installerExe := files[0]
 
 	slog.Info("starting upgrade with " + installerExe)
-	slog.Info("upgrade log file " + logFile)
+	slog.Info("upgrade log file " + UpgradeLogFile)
 
 	// When running in debug mode, we'll be "verbose" and let the installer pop up and prompt
 	installArgs := []string{
-		"/CLOSEAPPLICATIONS",             // Quit the tray app if it's still running
-		"/LOG=" + filepath.Base(logFile), // Only relative seems reliable, so set pwd
-		"/FORCECLOSEAPPLICATIONS",        // Force close the tray app - might be needed
+		"/CLOSEAPPLICATIONS",                    // Quit the tray app if it's still running
+		"/LOG=" + filepath.Base(UpgradeLogFile), // Only relative seems reliable, so set pwd
+		"/FORCECLOSEAPPLICATIONS",               // Force close the tray app - might be needed
 	}
 	// When we're not in debug mode, make the upgrade as quiet as possible (no GUI, no prompts)
 	// TODO - temporarily disable since we're pinning in debug mode for the preview
@@ -59,7 +53,7 @@ func DoUpgrade(cancel context.CancelFunc, done chan int) error {
 	}
 
 	slog.Debug(fmt.Sprintf("starting installer: %s %v", installerExe, installArgs))
-	os.Chdir(filepath.Dir(logFile)) //nolint:errcheck
+	os.Chdir(filepath.Dir(UpgradeLogFile)) //nolint:errcheck
 	cmd := exec.Command(installerExe, installArgs...)
 
 	if err := cmd.Start(); err != nil {

app/main.go

@@ -2,15 +2,11 @@ package main
 
 // Compile with the following to get rid of the cmd pop up on windows
 // go build -ldflags="-H windowsgui" .
-var (
-	AppName       string
-	CLIName       string
-	AppDir        string
-	AppDataDir    string
-	AppLogFile    string
-	ServerLogFile string
+
+import (
+	"github.com/ollama/ollama/app/lifecycle"
 )
 
 func main() {
-	run()
+	lifecycle.Run()
 }

app/ollama.iss

@@ -88,8 +88,8 @@ DialogFontSize=12
 [Files]
 Source: ".\app.exe"; DestDir: "{app}"; DestName: "{#MyAppExeName}" ; Flags: ignoreversion 64bit
 Source: "..\ollama.exe"; DestDir: "{app}"; Flags: ignoreversion 64bit
-Source: "..\dist\windows-amd64\*.dll"; DestDir: "{app}"; Flags: ignoreversion 64bit
-Source: "..\dist\windows-amd64\ollama_runners\*"; DestDir: "{app}\ollama_runners"; Flags: ignoreversion 64bit recursesubdirs
+Source: "..\dist\windows-{#ARCH}\*.dll"; DestDir: "{app}"; Flags: ignoreversion 64bit
+Source: "..\dist\windows-{#ARCH}\ollama_runners\*"; DestDir: "{app}\ollama_runners"; Flags: ignoreversion 64bit recursesubdirs
 Source: "..\dist\ollama_welcome.ps1"; DestDir: "{app}"; Flags: ignoreversion
 Source: ".\assets\app.ico"; DestDir: "{app}"; Flags: ignoreversion
 #if DirExists("..\dist\windows-amd64\rocm")

app/tray/tray_nonwindows.go

@@ -1,3 +1,5 @@
+//go:build !windows
+
 package tray
 
 import (

auth/auth.go

@@ -10,12 +10,44 @@ import (
 	"log/slog"
 	"os"
 	"path/filepath"
+	"strings"
 
 	"golang.org/x/crypto/ssh"
 )
 
 const defaultPrivateKey = "id_ed25519"
 
+func keyPath() (string, error) {
+	home, err := os.UserHomeDir()
+	if err != nil {
+		return "", err
+	}
+
+	return filepath.Join(home, ".ollama", defaultPrivateKey), nil
+}
+
+func GetPublicKey() (string, error) {
+	keyPath, err := keyPath()
+	if err != nil {
+		return "", err
+	}
+
+	privateKeyFile, err := os.ReadFile(keyPath)
+	if err != nil {
+		slog.Info(fmt.Sprintf("Failed to load private key: %v", err))
+		return "", err
+	}
+
+	privateKey, err := ssh.ParsePrivateKey(privateKeyFile)
+	if err != nil {
+		return "", err
+	}
+
+	publicKey := ssh.MarshalAuthorizedKey(privateKey.PublicKey())
+
+	return strings.TrimSpace(string(publicKey)), nil
+}
+
 func NewNonce(r io.Reader, length int) (string, error) {
 	nonce := make([]byte, length)
 	if _, err := io.ReadFull(r, nonce); err != nil {
@@ -26,13 +58,11 @@ func NewNonce(r io.Reader, length int) (string, error) {
 }
 
 func Sign(ctx context.Context, bts []byte) (string, error) {
-	home, err := os.UserHomeDir()
+	keyPath, err := keyPath()
 	if err != nil {
 		return "", err
 	}
 
-	keyPath := filepath.Join(home, ".ollama", defaultPrivateKey)
-
 	privateKeyFile, err := os.ReadFile(keyPath)
 	if err != nil {
 		slog.Info(fmt.Sprintf("Failed to load private key: %v", err))

cmd/cmd.go

@@ -32,10 +32,12 @@ import (
 	"golang.org/x/term"
 
 	"github.com/ollama/ollama/api"
+	"github.com/ollama/ollama/auth"
 	"github.com/ollama/ollama/format"
-	"github.com/ollama/ollama/parser"
 	"github.com/ollama/ollama/progress"
 	"github.com/ollama/ollama/server"
+	"github.com/ollama/ollama/types/errtypes"
+	"github.com/ollama/ollama/types/model"
 	"github.com/ollama/ollama/version"
 )
 
@@ -54,12 +56,13 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 	p := progress.NewProgress(os.Stderr)
 	defer p.Stop()
 
-	modelfile, err := os.ReadFile(filename)
+	f, err := os.Open(filename)
 	if err != nil {
 		return err
 	}
+	defer f.Close()
 
-	commands, err := parser.Parse(bytes.NewReader(modelfile))
+	modelfile, err := model.ParseFile(f)
 	if err != nil {
 		return err
 	}
@@ -73,10 +76,10 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 	spinner := progress.NewSpinner(status)
 	p.Add(status, spinner)
 
-	for _, c := range commands {
-		switch c.Name {
+	for i := range modelfile.Commands {
+		switch modelfile.Commands[i].Name {
 		case "model", "adapter":
-			path := c.Args
+			path := modelfile.Commands[i].Args
 			if path == "~" {
 				path = home
 			} else if strings.HasPrefix(path, "~/") {
@@ -88,7 +91,7 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 			}
 
 			fi, err := os.Stat(path)
-			if errors.Is(err, os.ErrNotExist) && c.Name == "model" {
+			if errors.Is(err, os.ErrNotExist) && modelfile.Commands[i].Name == "model" {
 				continue
 			} else if err != nil {
 				return err
@@ -111,13 +114,7 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 				return err
 			}
 
-			name := c.Name
-			if c.Name == "model" {
-				name = "from"
-			}
-
-			re := regexp.MustCompile(fmt.Sprintf(`(?im)^(%s)\s+%s\s*$`, name, c.Args))
-			modelfile = re.ReplaceAll(modelfile, []byte("$1 @"+digest))
+			modelfile.Commands[i].Args = "@" + digest
 		}
 	}
 
@@ -147,7 +144,7 @@ func CreateHandler(cmd *cobra.Command, args []string) error {
 
 	quantization, _ := cmd.Flags().GetString("quantization")
 
-	request := api.CreateRequest{Name: args[0], Modelfile: string(modelfile), Quantization: quantization}
+	request := api.CreateRequest{Name: args[0], Modelfile: modelfile.String(), Quantization: quantization}
 	if err := client.Create(cmd.Context(), &request, fn); err != nil {
 		return err
 	}
@@ -357,6 +354,47 @@ func RunHandler(cmd *cobra.Command, args []string) error {
 	return generateInteractive(cmd, opts)
 }
 
+func errFromUnknownKey(unknownKeyErr error) error {
+	// find SSH public key in the error message
+	sshKeyPattern := `ssh-\w+ [^\s"]+`
+	re := regexp.MustCompile(sshKeyPattern)
+	matches := re.FindStringSubmatch(unknownKeyErr.Error())
+
+	if len(matches) > 0 {
+		serverPubKey := matches[0]
+
+		localPubKey, err := auth.GetPublicKey()
+		if err != nil {
+			return unknownKeyErr
+		}
+
+		if runtime.GOOS == "linux" && serverPubKey != localPubKey {
+			// try the ollama service public key
+			svcPubKey, err := os.ReadFile("/usr/share/ollama/.ollama/id_ed25519.pub")
+			if err != nil {
+				return unknownKeyErr
+			}
+			localPubKey = strings.TrimSpace(string(svcPubKey))
+		}
+
+		// check if the returned public key matches the local public key, this prevents adding a remote key to the user's account
+		if serverPubKey != localPubKey {
+			return unknownKeyErr
+		}
+
+		var msg strings.Builder
+		msg.WriteString(unknownKeyErr.Error())
+		msg.WriteString("\n\nYour ollama key is:\n")
+		msg.WriteString(localPubKey)
+		msg.WriteString("\nAdd your key at:\n")
+		msg.WriteString("https://ollama.com/settings/keys")
+
+		return errors.New(msg.String())
+	}
+
+	return unknownKeyErr
+}
+
 func PushHandler(cmd *cobra.Command, args []string) error {
 	client, err := api.ClientFromEnvironment()
 	if err != nil {
@@ -404,6 +442,20 @@ func PushHandler(cmd *cobra.Command, args []string) error {
 
 	request := api.PushRequest{Name: args[0], Insecure: insecure}
 	if err := client.Push(cmd.Context(), &request, fn); err != nil {
+		if spinner != nil {
+			spinner.Stop()
+		}
+		if strings.Contains(err.Error(), "access denied") {
+			return errors.New("you are not authorized to push to this namespace, create the model under a namespace you own")
+		}
+		host := model.ParseName(args[0]).Host
+		isOllamaHost := strings.HasSuffix(host, ".ollama.ai") || strings.HasSuffix(host, ".ollama.com")
+		if strings.Contains(err.Error(), errtypes.UnknownOllamaKeyErrMsg) && isOllamaHost {
+			// the user has not added their ollama key to ollama.com
+			// re-throw an error with a more user-friendly message
+			return errFromUnknownKey(err)
+		}
+
 		return err
 	}
 
@@ -831,19 +883,17 @@ func generate(cmd *cobra.Command, opts runOptions) error {
 }
 
 func RunServer(cmd *cobra.Command, _ []string) error {
-	host, port, err := net.SplitHostPort(strings.Trim(os.Getenv("OLLAMA_HOST"), "\"'"))
+	// retrieve the OLLAMA_HOST environment variable
+	ollamaHost, err := api.GetOllamaHost()
 	if err != nil {
-		host, port = "127.0.0.1", "11434"
-		if ip := net.ParseIP(strings.Trim(os.Getenv("OLLAMA_HOST"), "[]")); ip != nil {
-			host = ip.String()
-		}
+		return err
 	}
 
 	if err := initializeKeypair(); err != nil {
 		return err
 	}
 
-	ln, err := net.Listen("tcp", net.JoinHostPort(host, port))
+	ln, err := net.Listen("tcp", net.JoinHostPort(ollamaHost.Host, ollamaHost.Port))
 	if err != nil {
 		return err
 	}
@@ -1069,7 +1119,7 @@ Environment Variables:
 		RunE:    ListHandler,
 	}
 	copyCmd := &cobra.Command{
-		Use:     "cp SOURCE TARGET",
+		Use:     "cp SOURCE DESTINATION",
 		Short:   "Copy a model",
 		Args:    cobra.ExactArgs(2),
 		PreRunE: checkServerHeartbeat,

cmd/interactive.go

@@ -94,6 +94,7 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 		fmt.Fprintln(os.Stderr, "  /show           Show model information")
 		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, "  /bye            Exit")
 		fmt.Fprintln(os.Stderr, "  /?, /help       Help for a command")
 		fmt.Fprintln(os.Stderr, "  /? shortcuts    Help for keyboard shortcuts")
@@ -280,6 +281,10 @@ func generateInteractive(cmd *cobra.Command, opts runOptions) error {
 			}
 			fmt.Printf("Created new model '%s'\n", args[1])
 			continue
+		case strings.HasPrefix(line, "/clear"):
+			opts.Messages = []api.Message{}
+			fmt.Println("Cleared session context")
+			continue
 		case strings.HasPrefix(line, "/set"):
 			args := strings.Fields(line)
 			if len(args) > 1 {

convert/convert.go

@@ -5,6 +5,7 @@ import (
 	"encoding/binary"
 	"encoding/json"
 	"fmt"
+	"io"
 	"log/slog"
 	"os"
 	"path/filepath"
@@ -47,7 +48,7 @@ type ByteOrder interface {
 type ModelArch interface {
 	GetTensors() error
 	LoadVocab() error
-	WriteGGUF() (string, error)
+	WriteGGUF(io.WriteSeeker) error
 }
 
 type ModelFormat interface {

convert/gemma.go

@@ -94,7 +94,7 @@ func (m *GemmaModel) LoadVocab() error {
 	return nil
 }
 
-func (m *GemmaModel) WriteGGUF() (string, error) {
+func (m *GemmaModel) WriteGGUF(ws io.WriteSeeker) error {
 	kv := llm.KV{
 		"general.architecture":                   "gemma",
 		"general.name":                           m.Name,
@@ -122,16 +122,5 @@ func (m *GemmaModel) WriteGGUF() (string, error) {
 		"tokenizer.ggml.add_eos_token":    false,
 	}
 
-	f, err := os.CreateTemp("", "ollama-gguf")
-	if err != nil {
-		return "", err
-	}
-	defer f.Close()
-
-	mod := llm.NewGGUFV3(m.Params.ByteOrder)
-	if err := mod.Encode(f, kv, m.Tensors); err != nil {
-		return "", err
-	}
-
-	return f.Name(), nil
+	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(ws, kv, m.Tensors)
 }

convert/llama.go

@@ -132,7 +132,7 @@ func (m *LlamaModel) LoadVocab() error {
 	return nil
 }
 
-func (m *LlamaModel) WriteGGUF() (string, error) {
+func (m *LlamaModel) WriteGGUF(ws io.WriteSeeker) error {
 	kv := llm.KV{
 		"general.architecture":                   "llama",
 		"general.name":                           m.Name,
@@ -161,16 +161,9 @@ func (m *LlamaModel) WriteGGUF() (string, error) {
 
 	f, err := os.CreateTemp("", "ollama-gguf")
 	if err != nil {
-		return "", err
+		return err
 	}
 	defer f.Close()
 
-	mod := llm.NewGGUFV3(m.Params.ByteOrder)
-	if err := mod.Encode(f, kv, m.Tensors); err != nil {
-		return "", err
-	}
-
-	slog.Debug(fmt.Sprintf("gguf file = %s", f.Name()))
-
-	return f.Name(), nil
+	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(f, kv, m.Tensors)
 }

convert/mistral.go

@@ -132,7 +132,7 @@ func (m *MistralModel) LoadVocab() error {
 	return nil
 }
 
-func (m *MistralModel) WriteGGUF() (string, error) {
+func (m *MistralModel) WriteGGUF(ws io.WriteSeeker) error {
 	kv := llm.KV{
 		"general.architecture":                   "llama",
 		"general.name":                           m.Name,
@@ -158,16 +158,5 @@ func (m *MistralModel) WriteGGUF() (string, error) {
 		"tokenizer.ggml.unknown_token_id": uint32(0),
 	}
 
-	f, err := os.CreateTemp("", "ollama-gguf")
-	if err != nil {
-		return "", err
-	}
-	defer f.Close()
-
-	mod := llm.NewGGUFV3(m.Params.ByteOrder)
-	if err := mod.Encode(f, kv, m.Tensors); err != nil {
-		return "", err
-	}
-
-	return f.Name(), nil
+	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(ws, kv, m.Tensors)
 }

convert/mixtral.go

@@ -1,7 +1,7 @@
 package convert
 
 import (
-	"os"
+	"io"
 	"regexp"
 
 	"github.com/ollama/ollama/llm"
@@ -47,7 +47,7 @@ func (m *MixtralModel) LoadVocab() error {
 	return nil
 }
 
-func (m *MixtralModel) WriteGGUF() (string, error) {
+func (m *MixtralModel) WriteGGUF(ws io.WriteSeeker) error {
 	kv := llm.KV{
 		"general.architecture":          "llama",
 		"general.name":                  m.Name,
@@ -81,16 +81,5 @@ func (m *MixtralModel) WriteGGUF() (string, error) {
 		"tokenizer.ggml.add_eos_token":    false,
 	}
 
-	f, err := os.CreateTemp("", "ollama-gguf")
-	if err != nil {
-		return "", err
-	}
-	defer f.Close()
-
-	mod := llm.NewGGUFV3(m.Params.ByteOrder)
-	if err := mod.Encode(f, kv, m.Tensors); err != nil {
-		return "", err
-	}
-
-	return f.Name(), nil
+	return llm.NewGGUFV3(m.Params.ByteOrder).Encode(ws, kv, m.Tensors)
 }

docs/api.md

@@ -17,7 +17,7 @@
 
 ### Model names
 
-Model names follow a `model:tag` format, where `model` can have an optional namespace such as `example/model`. Some examples are `orca-mini:3b-q4_1` and `llama2:70b`. The tag is optional and, if not provided, will default to `latest`. The tag is used to identify a specific version.
+Model names follow a `model:tag` format, where `model` can have an optional namespace such as `example/model`. Some examples are `orca-mini:3b-q4_1` and `llama3:70b`. The tag is optional and, if not provided, will default to `latest`. The tag is used to identify a specific version.
 
 ### Durations
 
@@ -66,7 +66,7 @@ Enable JSON mode by setting the `format` parameter to `json`. This will structur
 
 ```shell
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2",
+  "model": "llama3",
   "prompt": "Why is the sky blue?"
 }'
 ```
@@ -77,7 +77,7 @@ A stream of JSON objects is returned:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T08:52:19.385406455-07:00",
   "response": "The",
   "done": false
@@ -99,7 +99,7 @@ To calculate how fast the response is generated in tokens per second (token/s),
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T19:22:45.499127Z",
   "response": "",
   "done": true,
@@ -121,7 +121,7 @@ A response can be received in one reply when streaming is off.
 
 ```shell
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2",
+  "model": "llama3",
   "prompt": "Why is the sky blue?",
   "stream": false
 }'
@@ -133,7 +133,7 @@ If `stream` is set to `false`, the response will be a single JSON object:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T19:22:45.499127Z",
   "response": "The sky is blue because it is the color of the sky.",
   "done": true,
@@ -155,7 +155,7 @@ If `stream` is set to `false`, the response will be a single JSON object:
 
 ```shell
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2",
+  "model": "llama3",
   "prompt": "What color is the sky at different times of the day? Respond using JSON",
   "format": "json",
   "stream": false
@@ -166,7 +166,7 @@ curl http://localhost:11434/api/generate -d '{
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-11-09T21:07:55.186497Z",
   "response": "{\n\"morning\": {\n\"color\": \"blue\"\n},\n\"noon\": {\n\"color\": \"blue-gray\"\n},\n\"afternoon\": {\n\"color\": \"warm gray\"\n},\n\"evening\": {\n\"color\": \"orange\"\n}\n}\n",
   "done": true,
@@ -289,7 +289,7 @@ If you want to set custom options for the model at runtime rather than in the Mo
 
 ```shell
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2",
+  "model": "llama3",
   "prompt": "Why is the sky blue?",
   "stream": false,
   "options": {
@@ -332,7 +332,7 @@ curl http://localhost:11434/api/generate -d '{
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T19:22:45.499127Z",
   "response": "The sky is blue because it is the color of the sky.",
   "done": true,
@@ -354,7 +354,7 @@ If an empty prompt is provided, the model will be loaded into memory.
 
 ```shell
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2"
+  "model": "llama3"
 }'
 ```
 
@@ -364,7 +364,7 @@ A single JSON object is returned:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-12-18T19:52:07.071755Z",
   "response": "",
   "done": true
@@ -407,7 +407,7 @@ Send a chat message with a streaming response.
 
 ```shell
 curl http://localhost:11434/api/chat -d '{
-  "model": "llama2",
+  "model": "llama3",
   "messages": [
     {
       "role": "user",
@@ -423,7 +423,7 @@ A stream of JSON objects is returned:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T08:52:19.385406455-07:00",
   "message": {
     "role": "assistant",
@@ -438,7 +438,7 @@ Final response:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T19:22:45.499127Z",
   "done": true,
   "total_duration": 4883583458,
@@ -456,7 +456,7 @@ Final response:
 
 ```shell
 curl http://localhost:11434/api/chat -d '{
-  "model": "llama2",
+  "model": "llama3",
   "messages": [
     {
       "role": "user",
@@ -471,7 +471,7 @@ curl http://localhost:11434/api/chat -d '{
 
 ```json
 {
-  "model": "registry.ollama.ai/library/llama2:latest",
+  "model": "registry.ollama.ai/library/llama3:latest",
   "created_at": "2023-12-12T14:13:43.416799Z",
   "message": {
     "role": "assistant",
@@ -495,7 +495,7 @@ Send a chat message with a conversation history. You can use this same approach
 
 ```shell
 curl http://localhost:11434/api/chat -d '{
-  "model": "llama2",
+  "model": "llama3",
   "messages": [
     {
       "role": "user",
@@ -519,7 +519,7 @@ A stream of JSON objects is returned:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T08:52:19.385406455-07:00",
   "message": {
     "role": "assistant",
@@ -533,7 +533,7 @@ Final response:
 
 ```json
 {
-  "model": "llama2",
+  "model": "llama3",
   "created_at": "2023-08-04T19:22:45.499127Z",
   "done": true,
   "total_duration": 8113331500,
@@ -591,7 +591,7 @@ curl http://localhost:11434/api/chat -d '{
 
 ```shell
 curl http://localhost:11434/api/chat -d '{
-  "model": "llama2",
+  "model": "llama3",
   "messages": [
     {
       "role": "user",
@@ -609,7 +609,7 @@ curl http://localhost:11434/api/chat -d '{
 
 ```json
 {
-  "model": "registry.ollama.ai/library/llama2:latest",
+  "model": "registry.ollama.ai/library/llama3:latest",
   "created_at": "2023-12-12T14:13:43.416799Z",
   "message": {
     "role": "assistant",
@@ -651,7 +651,7 @@ Create a new model from a `Modelfile`.
 ```shell
 curl http://localhost:11434/api/create -d '{
   "name": "mario",
-  "modelfile": "FROM llama2\nSYSTEM You are mario from Super Mario Bros."
+  "modelfile": "FROM llama3\nSYSTEM You are mario from Super Mario Bros."
 }'
 ```
 
@@ -758,7 +758,7 @@ A single JSON object will be returned.
       }
     },
     {
-      "name": "llama2:latest",
+      "name": "llama3:latest",
       "modified_at": "2023-12-07T09:32:18.757212583-08:00",
       "size": 3825819519,
       "digest": "fe938a131f40e6f6d40083c9f0f430a515233eb2edaa6d72eb85c50d64f2300e",
@@ -792,7 +792,7 @@ Show information about a model including details, modelfile, template, parameter
 
 ```shell
 curl http://localhost:11434/api/show -d '{
-  "name": "llama2"
+  "name": "llama3"
 }'
 ```
 
@@ -827,8 +827,8 @@ Copy a model. Creates a model with another name from an existing model.
 
 ```shell
 curl http://localhost:11434/api/copy -d '{
-  "source": "llama2",
-  "destination": "llama2-backup"
+  "source": "llama3",
+  "destination": "llama3-backup"
 }'
 ```
 
@@ -854,7 +854,7 @@ Delete a model and its data.
 
 ```shell
 curl -X DELETE http://localhost:11434/api/delete -d '{
-  "name": "llama2:13b"
+  "name": "llama3:13b"
 }'
 ```
 
@@ -882,7 +882,7 @@ Download a model from the ollama library. Cancelled pulls are resumed from where
 
 ```shell
 curl http://localhost:11434/api/pull -d '{
-  "name": "llama2"
+  "name": "llama3"
 }'
 ```
 

docs/development.md

@@ -51,7 +51,7 @@ Typically the build scripts will auto-detect CUDA, however, if your Linux distro
 or installation approach uses unusual paths, you can specify the location by
 specifying an environment variable `CUDA_LIB_DIR` to the location of the shared
 libraries, and `CUDACXX` to the location of the nvcc compiler. You can customize
-set set of target CUDA architectues by setting `CMAKE_CUDA_ARCHITECTURES` (e.g. "50;60;70")
+a set of target CUDA architectures by setting `CMAKE_CUDA_ARCHITECTURES` (e.g. "50;60;70")
 
 Then generate dependencies:
 
@@ -142,4 +142,4 @@ In addition to the common Windows development tools described above, install AMD
 - [AMD HIP](https://www.amd.com/en/developer/resources/rocm-hub/hip-sdk.html)
 - [Strawberry Perl](https://strawberryperl.com/)
 
-Lastly, add `ninja.exe` included with MSVC to the system path (e.g. `C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\Common7\IDE\CommonExtensions\Microsoft\CMake\Ninja`).
+Lastly, add `ninja.exe` included with MSVC to the system path (e.g. `C:\Program Files (x86)\Microsoft Visual Studio\2019\Community\Common7\IDE\CommonExtensions\Microsoft\CMake\Ninja`).

docs/faq.md

@@ -32,7 +32,7 @@ When using the API, specify the `num_ctx` parameter:
 
 ```
 curl http://localhost:11434/api/generate -d '{
-  "model": "llama2",
+  "model": "llama3",
   "prompt": "Why is the sky blue?",
   "options": {
     "num_ctx": 4096
@@ -88,9 +88,9 @@ On windows, Ollama inherits your user and system environment variables.
 
 3. Edit or create New variable(s) for your user account for `OLLAMA_HOST`, `OLLAMA_MODELS`, etc.
 
-4. Click OK/Apply to save 
+4. Click OK/Apply to save
 
-5. Run `ollama` from a new terminal window 
+5. Run `ollama` from a new terminal window
 
 
 ## How can I expose Ollama on my network?
@@ -221,12 +221,12 @@ The `keep_alive` parameter can be set to:
 
 For example, to preload a model and leave it in memory use:
 ```shell
-curl http://localhost:11434/api/generate -d '{"model": "llama2", "keep_alive": -1}'
+curl http://localhost:11434/api/generate -d '{"model": "llama3", "keep_alive": -1}'
 ```
 
 To unload the model and free up memory use:
 ```shell
-curl http://localhost:11434/api/generate -d '{"model": "llama2", "keep_alive": 0}'
+curl http://localhost:11434/api/generate -d '{"model": "llama3", "keep_alive": 0}'
 ```
 
 Alternatively, you can change the amount of time all models are loaded into memory by setting the `OLLAMA_KEEP_ALIVE` environment variable when starting the Ollama server. The `OLLAMA_KEEP_ALIVE` variable uses the same parameter types as the `keep_alive` parameter types mentioned above. Refer to section explaining [how to configure the Ollama server](#how-do-i-configure-ollama-server) to correctly set the environment variable.

docs/modelfile.md

@@ -10,7 +10,7 @@ A model file is the blueprint to create and share models with Ollama.
 - [Examples](#examples)
 - [Instructions](#instructions)
   - [FROM (Required)](#from-required)
-    - [Build from llama2](#build-from-llama2)
+    - [Build from llama3](#build-from-llama3)
     - [Build from a bin file](#build-from-a-bin-file)
   - [PARAMETER](#parameter)
     - [Valid Parameters and Values](#valid-parameters-and-values)
@@ -48,7 +48,7 @@ INSTRUCTION arguments
 An example of a `Modelfile` creating a mario blueprint:
 
 ```modelfile
-FROM llama2
+FROM llama3
 # sets the temperature to 1 [higher is more creative, lower is more coherent]
 PARAMETER temperature 1
 # sets the context window size to 4096, this controls how many tokens the LLM can use as context to generate the next token
@@ -67,33 +67,25 @@ To use this:
 
 More examples are available in the [examples directory](../examples).
 
-### `Modelfile`s in [ollama.com/library][1]
-
-There are two ways to view `Modelfile`s underlying the models in [ollama.com/library][1]:
-
-- Option 1: view a details page from a model's tags page:
-  1.  Go to a particular model's tags (e.g. https://ollama.com/library/llama2/tags)
-  2.  Click on a tag (e.g. https://ollama.com/library/llama2:13b)
-  3.  Scroll down to "Layers"
-      - Note: if the [`FROM` instruction](#from-required) is not present,
-        it means the model was created from a local file
-- Option 2: use `ollama show` to print the `Modelfile` for any local models like so:
+To view the Modelfile of a given model, use the `ollama show --modelfile` command.
 
   ```bash
-  > ollama show --modelfile llama2:13b
+  > ollama show --modelfile llama3
   # Modelfile generated by "ollama show"
   # To build a new Modelfile based on this one, replace the FROM line with:
-  # FROM llama2:13b
+  # FROM llama3:latest
+  FROM /Users/pdevine/.ollama/models/blobs/sha256-00e1317cbf74d901080d7100f57580ba8dd8de57203072dc6f668324ba545f29
+  TEMPLATE """{{ if .System }}<|start_header_id|>system<|end_header_id|>
 
-  FROM /root/.ollama/models/blobs/sha256:123abc
-  TEMPLATE """[INST] {{ if .System }}<<SYS>>{{ .System }}<</SYS>>
+  {{ .System }}<|eot_id|>{{ end }}{{ if .Prompt }}<|start_header_id|>user<|end_header_id|>
 
-  {{ end }}{{ .Prompt }} [/INST] """
-  SYSTEM """"""
-  PARAMETER stop [INST]
-  PARAMETER stop [/INST]
-  PARAMETER stop <<SYS>>
-  PARAMETER stop <</SYS>>
+  {{ .Prompt }}<|eot_id|>{{ end }}<|start_header_id|>assistant<|end_header_id|>
+
+  {{ .Response }}<|eot_id|>"""
+  PARAMETER stop "<|start_header_id|>"
+  PARAMETER stop "<|end_header_id|>"
+  PARAMETER stop "<|eot_id|>"
+  PARAMETER stop "<|reserved_special_token"
   ```
 
 ## Instructions
@@ -106,10 +98,10 @@ The `FROM` instruction defines the base model to use when creating a model.
 FROM <model name>:<tag>
 ```
 
-#### Build from llama2
+#### Build from llama3
 
 ```modelfile
-FROM llama2
+FROM llama3
 ```
 
 A list of available base models:

docs/openai.md

@@ -25,7 +25,7 @@ chat_completion = client.chat.completions.create(
             'content': 'Say this is a test',
         }
     ],
-    model='llama2',
+    model='llama3',
 )
 ```
 
@@ -43,7 +43,7 @@ const openai = new OpenAI({
 
 const chatCompletion = await openai.chat.completions.create({
   messages: [{ role: 'user', content: 'Say this is a test' }],
-  model: 'llama2',
+  model: 'llama3',
 })
 ```
 
@@ -53,7 +53,7 @@ const chatCompletion = await openai.chat.completions.create({
 curl http://localhost:11434/v1/chat/completions \
     -H "Content-Type: application/json" \
     -d '{
-        "model": "llama2",
+        "model": "llama3",
         "messages": [
             {
                 "role": "system",
@@ -113,7 +113,7 @@ curl http://localhost:11434/v1/chat/completions \
 Before using a model, pull it locally `ollama pull`:
 
 ```shell
-ollama pull llama2
+ollama pull llama3
 ```
 
 ### Default model names
@@ -121,7 +121,7 @@ ollama pull llama2
 For tooling that relies on default OpenAI model names such as `gpt-3.5-turbo`, use `ollama cp` to copy an existing model name to a temporary name:
 
 ```
-ollama cp llama2 gpt-3.5-turbo
+ollama cp llama3 gpt-3.5-turbo
 ```
 
 Afterwards, this new model name can be specified the `model` field:

docs/tutorials/langchainjs.md

@@ -15,7 +15,7 @@ import { Ollama } from "langchain/llms/ollama";
 
 const ollama = new Ollama({
   baseUrl: "http://localhost:11434",
-  model: "llama2",
+  model: "llama3",
 });
 
 const answer = await ollama.invoke(`why is the sky blue?`);
@@ -23,10 +23,10 @@ const answer = await ollama.invoke(`why is the sky blue?`);
 console.log(answer);
 ```
 
-That will get us the same thing as if we ran `ollama run llama2 "why is the sky blue"` in the terminal. But we want to load a document from the web to ask a question against. **Cheerio** is a great library for ingesting a webpage, and **LangChain** uses it in their **CheerioWebBaseLoader**. So let's install **Cheerio** and build that part of the app.
+That will get us the same thing as if we ran `ollama run llama3 "why is the sky blue"` in the terminal. But we want to load a document from the web to ask a question against. **Cheerio** is a great library for ingesting a webpage, and **LangChain** uses it in their **CheerioWebBaseLoader**. So let's install **Cheerio** and build that part of the app.
 
 ```bash
-npm install cheerio 
+npm install cheerio
 ```
 
 ```javascript

docs/tutorials/langchainpy.md

@@ -17,10 +17,12 @@ Let's start by asking a simple question that we can get an answer to from the **
 Then we can create a model and ask the question:
 
 ```python
-from langchain.llms import Ollama
-ollama = Ollama(base_url='http://localhost:11434',
-model="llama2")
-print(ollama("why is the sky blue"))
+from langchain_community.llms import Ollama
+ollama = Ollama(
+    base_url='http://localhost:11434',
+    model="llama3"
+)
+print(ollama.invoke("why is the sky blue"))
 ```
 
 Notice that we are defining the model and the base URL for Ollama.

docs/windows.md

@@ -1,47 +1,48 @@
-# Ollama Windows Preview
-
-Welcome to the Ollama Windows preview.
-
-No more WSL required!
-
-Ollama now runs as a native Windows application, including NVIDIA and AMD Radeon GPU support.
-After installing Ollama Windows Preview, Ollama will run in the background and
-the `ollama` command line is available in `cmd`, `powershell` or your favorite
-terminal application. As usual the Ollama [api](./api.md) will be served on
-`http://localhost:11434`.
-
-As this is a preview release, you should expect a few bugs here and there.  If
-you run into a problem you can reach out on
-[Discord](https://discord.gg/ollama), or file an 
-[issue](https://github.com/ollama/ollama/issues).
-Logs will often be helpful in diagnosing the problem (see
-[Troubleshooting](#troubleshooting) below)
-
-## System Requirements
-
-* Windows 10 or newer, Home or Pro
-* NVIDIA 452.39 or newer Drivers if you have an NVIDIA card
-* AMD Radeon Driver https://www.amd.com/en/support if you have a Radeon card
-
-## API Access
-
-Here's a quick example showing API access from `powershell`
-```powershell
-(Invoke-WebRequest -method POST -Body '{"model":"llama2", "prompt":"Why is the sky blue?", "stream": false}' -uri http://localhost:11434/api/generate ).Content | ConvertFrom-json
-```
-
-## Troubleshooting
-
-While we're in preview, `OLLAMA_DEBUG` is always enabled, which adds
-a "view logs" menu item to the app, and increses logging for the GUI app and
-server.
-
-Ollama on Windows stores files in a few different locations.  You can view them in
-the explorer window by hitting `<cmd>+R` and type in:
-- `explorer %LOCALAPPDATA%\Ollama` contains logs, and downloaded updates
-    - *app.log* contains logs from the GUI application
-    - *server.log* contains the server logs
-    - *upgrade.log* contains log output for upgrades
-- `explorer %LOCALAPPDATA%\Programs\Ollama` contains the binaries (The installer adds this to your user PATH)
-- `explorer %HOMEPATH%\.ollama` contains models and configuration
-- `explorer %TEMP%` contains temporary executable files in one or more `ollama*` directories
+<<<<<<< HEAD
+# Ollama Windows Preview
+
+Welcome to the Ollama Windows preview.
+
+No more WSL required!
+
+Ollama now runs as a native Windows application, including NVIDIA and AMD Radeon GPU support.
+After installing Ollama Windows Preview, Ollama will run in the background and
+the `ollama` command line is available in `cmd`, `powershell` or your favorite
+terminal application. As usual the Ollama [api](./api.md) will be served on
+`http://localhost:11434`.
+
+As this is a preview release, you should expect a few bugs here and there.  If
+you run into a problem you can reach out on
+[Discord](https://discord.gg/ollama), or file an
+[issue](https://github.com/ollama/ollama/issues).
+Logs will often be helpful in diagnosing the problem (see
+[Troubleshooting](#troubleshooting) below)
+
+## System Requirements
+
+* Windows 10 or newer, Home or Pro
+* NVIDIA 452.39 or newer Drivers if you have an NVIDIA card
+* AMD Radeon Driver https://www.amd.com/en/support if you have a Radeon card
+
+## API Access
+
+Here's a quick example showing API access from `powershell`
+```powershell
+(Invoke-WebRequest -method POST -Body '{"model":"llama3", "prompt":"Why is the sky blue?", "stream": false}' -uri http://localhost:11434/api/generate ).Content | ConvertFrom-json
+```
+
+## Troubleshooting
+
+While we're in preview, `OLLAMA_DEBUG` is always enabled, which adds
+a "view logs" menu item to the app, and increses logging for the GUI app and
+server.
+
+Ollama on Windows stores files in a few different locations.  You can view them in
+the explorer window by hitting `<cmd>+R` and type in:
+- `explorer %LOCALAPPDATA%\Ollama` contains logs, and downloaded updates
+    - *app.log* contains logs from the GUI application
+    - *server.log* contains the server logs
+    - *upgrade.log* contains log output for upgrades
+- `explorer %LOCALAPPDATA%\Programs\Ollama` contains the binaries (The installer adds this to your user PATH)
+- `explorer %HOMEPATH%\.ollama` contains models and configuration
+- `explorer %TEMP%` contains temporary executable files in one or more `ollama*` directories

examples/bash-comparemodels/README.md

@@ -2,7 +2,7 @@
 
 When calling `ollama`, you can pass it a file to run all the prompts in the file, one after the other:
 
-`ollama run llama2 < sourcequestions.txt`
+`ollama run llama3 < sourcequestions.txt`
 
 This concept is used in the following example.
 

examples/go-chat/main.go

@@ -35,7 +35,7 @@ func main() {
 
 	ctx := context.Background()
 	req := &api.ChatRequest{
-		Model:    "llama2",
+		Model:    "llama3",
 		Messages: messages,
 	}
 

examples/langchain-python-rag-document/main.py

@@ -40,9 +40,9 @@ while True:
         continue
 
     # Prompt
-    template = """Use the following pieces of context to answer the question at the end. 
-    If you don't know the answer, just say that you don't know, don't try to make up an answer. 
-    Use three sentences maximum and keep the answer as concise as possible. 
+    template = """Use the following pieces of context to answer the question at the end.
+    If you don't know the answer, just say that you don't know, don't try to make up an answer.
+    Use three sentences maximum and keep the answer as concise as possible.
     {context}
     Question: {question}
     Helpful Answer:"""
@@ -51,11 +51,11 @@ while True:
         template=template,
     )
 
-    llm = Ollama(model="llama2:13b", callback_manager=CallbackManager([StreamingStdOutCallbackHandler()]))
+    llm = Ollama(model="llama3:8b", callback_manager=CallbackManager([StreamingStdOutCallbackHandler()]))
     qa_chain = RetrievalQA.from_chain_type(
         llm,
         retriever=vectorstore.as_retriever(),
         chain_type_kwargs={"prompt": QA_CHAIN_PROMPT},
     )
 
-    result = qa_chain({"query": query})
+    result = qa_chain({"query": query})

examples/langchain-python-simple/README.md

@@ -4,10 +4,10 @@ This example is a basic "hello world" of using LangChain with Ollama.
 
 ## Running the Example
 
-1. Ensure you have the `llama2` model installed:
+1. Ensure you have the `llama3` model installed:
 
    ```bash
-   ollama pull llama2
+   ollama pull llama3
    ```
 
 2. Install the Python Requirements.
@@ -21,4 +21,3 @@ This example is a basic "hello world" of using LangChain with Ollama.
    ```bash
    python main.py
    ```
-  

examples/langchain-python-simple/main.py

@@ -1,6 +1,6 @@
 from langchain.llms import Ollama
 
 input = input("What is your question?")
-llm = Ollama(model="llama2")
+llm = Ollama(model="llama3")
 res = llm.predict(input)
 print (res)

examples/modelfile-mario/Modelfile

@@ -1,4 +1,4 @@
-FROM llama2
+FROM llama3
 PARAMETER temperature 1
 SYSTEM """
 You are Mario from super mario bros, acting as an assistant.

examples/modelfile-mario/readme.md

@@ -2,12 +2,12 @@
 
 # Example character: Mario
 
-This example shows how to create a basic character using Llama2 as the base model.
+This example shows how to create a basic character using Llama3 as the base model.
 
 To run this example:
 
 1. Download the Modelfile
-2. `ollama pull llama2` to get the base model used in the model file.
+2. `ollama pull llama3` to get the base model used in the model file.
 3. `ollama create NAME -f ./Modelfile`
 4. `ollama run NAME`
 
@@ -18,7 +18,7 @@ Ask it some questions like "Who are you?" or "Is Peach in trouble again?"
 What the model file looks like:
 
 ```
-FROM llama2
+FROM llama3
 PARAMETER temperature 1
 SYSTEM """
 You are Mario from Super Mario Bros, acting as an assistant.

examples/python-json-datagenerator/predefinedschema.py

@@ -2,16 +2,16 @@ import requests
 import json
 import random
 
-model = "llama2"
+model = "llama3"
 template = {
-  "firstName": "", 
-  "lastName": "", 
+  "firstName": "",
+  "lastName": "",
   "address": {
-    "street": "", 
-    "city": "", 
-    "state": "", 
+    "street": "",
+    "city": "",
+    "state": "",
     "zipCode": ""
-  }, 
+  },
   "phoneNumber": ""
 }
 

examples/python-json-datagenerator/randomaddresses.py

@@ -12,7 +12,7 @@ countries = [
     "France",
 ]
 country = random.choice(countries)
-model = "llama2"
+model = "llama3"
 
 prompt = f"generate one realistically believable sample data set of a persons first name, last name, address in {country}, and phone number. Do not use common names. Respond using JSON. Key names should have no backslashes, values should use plain ascii with no special characters."
 

examples/python-json-datagenerator/readme.md

@@ -6,10 +6,10 @@ There are two python scripts in this example. `randomaddresses.py` generates ran
 
 ## Running the Example
 
-1. Ensure you have the `llama2` model installed:
+1. Ensure you have the `llama3` model installed:
 
    ```bash
-   ollama pull llama2
+   ollama pull llama3
    ```
 
 2. Install the Python Requirements.

examples/python-simplechat/client.py

@@ -2,7 +2,7 @@ import json
 import requests
 
 # NOTE: ollama must be running for this to work, start the ollama app or run `ollama serve`
-model = "llama2"  # TODO: update this for whatever model you wish to use
+model = "llama3"  # TODO: update this for whatever model you wish to use
 
 
 def chat(messages):

examples/python-simplechat/readme.md

@@ -4,10 +4,10 @@ The **chat** endpoint is one of two ways to generate text from an LLM with Ollam
 
 ## Running the Example
 
-1. Ensure you have the `llama2` model installed:
+1. Ensure you have the `llama3` model installed:
 
    ```bash
-   ollama pull llama2
+   ollama pull llama3
    ```
 
 2. Install the Python Requirements.

examples/typescript-mentors/README.md

@@ -4,10 +4,10 @@ This example demonstrates how one would create a set of 'mentors' you can have a
 
 ## Usage
 
-1. Add llama2 to have the mentors ask your questions:
+1. Add llama3 to have the mentors ask your questions:
 
    ```bash
-   ollama pull llama2
+   ollama pull llama3
    ```
 
 2. Install prerequisites:

examples/typescript-mentors/character-generator.ts

@@ -15,7 +15,7 @@ async function characterGenerator() {
   ollama.setModel("stablebeluga2:70b-q4_K_M");
   const bio = await ollama.generate(`create a bio of ${character} in a single long paragraph. Instead of saying '${character} is...' or '${character} was...' use language like 'You are...' or 'You were...'. Then create a paragraph describing the speaking mannerisms and style of ${character}. Don't include anything about how ${character} looked or what they sounded like, just focus on the words they said. Instead of saying '${character} would say...' use language like 'You should say...'. If you use quotes, always use single quotes instead of double quotes. If there are any specific words or phrases you used a lot, show how you used them. `);
 
-  const thecontents = `FROM llama2\nSYSTEM """\n${bio.response.replace(/(\r\n|\n|\r)/gm, " ").replace('would', 'should')} All answers to questions should be related back to what you are most known for.\n"""`;
+  const thecontents = `FROM llama3\nSYSTEM """\n${bio.response.replace(/(\r\n|\n|\r)/gm, " ").replace('would', 'should')} All answers to questions should be related back to what you are most known for.\n"""`;
 
   fs.writeFile(path.join(directory, 'Modelfile'), thecontents, (err: any) => {
     if (err) throw err;
@@ -23,4 +23,4 @@ async function characterGenerator() {
   });
 }
 
-characterGenerator();
+characterGenerator();

examples/typescript-simplechat/client.ts

@@ -1,6 +1,6 @@
 import * as readline from "readline";
 
-const model = "llama2";
+const model = "llama3";
 type Message = {
   role: "assistant" | "user" | "system";
   content: string;
@@ -74,4 +74,4 @@ async function main() {
 
 }
 
-main();
+main();

gpu/assets.go

@@ -40,7 +40,7 @@ func PayloadsDir() (string, error) {
 			}
 
 			var paths []string
-			for _, root := range []string{appExe, cwd} {
+			for _, root := range []string{filepath.Dir(appExe), cwd} {
 				paths = append(paths,
 					filepath.Join(root),
 					filepath.Join(root, "windows-"+runtime.GOARCH),

gpu/gpu_darwin.go

@@ -1,3 +1,5 @@
+//go:build darwin
+
 package gpu
 
 /*
@@ -8,6 +10,12 @@ package gpu
 import "C"
 import (
 	"runtime"
+
+	"github.com/ollama/ollama/format"
+)
+
+const (
+	metalMinimumMemory = 512 * format.MebiByte
 )
 
 func GetGPUInfo() GpuInfoList {
@@ -30,7 +38,7 @@ func GetGPUInfo() GpuInfoList {
 	// TODO is there a way to gather actual allocated video memory? (currentAllocatedSize doesn't work)
 	info.FreeMemory = info.TotalMemory
 
-	info.MinimumMemory = 0
+	info.MinimumMemory = metalMinimumMemory
 	return []GpuInfo{info}
 }
 

integration/utils_test.go

@@ -107,7 +107,7 @@ func startServer(ctx context.Context, ollamaHost string) error {
 
 	if tmp := os.Getenv("OLLAMA_HOST"); tmp != ollamaHost {
 		slog.Info("setting env", "OLLAMA_HOST", ollamaHost)
-		os.Setenv("OLLAMA_HOST", ollamaHost)
+		t.Setenv("OLLAMA_HOST", ollamaHost)
 	}
 
 	slog.Info("starting server", "url", ollamaHost)

llm/ext_server/server.cpp

@@ -1032,7 +1032,7 @@ struct llama_server_context
             slot.has_next_token = false;
         }
 
-        if (!slot.cache_tokens.empty() && result.tok == llama_token_eos(model))
+        if (!slot.cache_tokens.empty() && llama_token_is_eog(model, result.tok))
         {
             slot.stopped_eos = true;
             slot.has_next_token = false;
@@ -1144,12 +1144,15 @@ struct llama_server_context
 
         res.result_json = json
         {
-            {"content",    tkn.text_to_send},
             {"stop",       false},
             {"slot_id",    slot.id},
             {"multimodal", multimodal}
         };
 
+        if (!llama_token_is_eog(model, tkn.tok)) {
+            res.result_json["content"] = tkn.text_to_send;
+        }
+
         if (slot.sparams.n_probs > 0)
         {
             std::vector<completion_token_output> probs_output = {};
@@ -2644,18 +2647,18 @@ static void server_params_parse(int argc, char **argv, server_params &sparams,
             if (strncmp(sep, "int:", 4) == 0) {
                 sep += 4;
                 kvo.tag = LLAMA_KV_OVERRIDE_TYPE_INT;
-                kvo.int_value = std::atol(sep);
+                kvo.val_i64 = std::atol(sep);
             } else if (strncmp(sep, "float:", 6) == 0) {
                 sep += 6;
                 kvo.tag = LLAMA_KV_OVERRIDE_TYPE_FLOAT;
-                kvo.float_value = std::atof(sep);
+                kvo.val_f64 = std::atof(sep);
             } else if (strncmp(sep, "bool:", 5) == 0) {
                 sep += 5;
                 kvo.tag = LLAMA_KV_OVERRIDE_TYPE_BOOL;
                 if (std::strcmp(sep, "true") == 0) {
-                    kvo.bool_value = true;
+                    kvo.val_bool = true;
                 } else if (std::strcmp(sep, "false") == 0) {
-                    kvo.bool_value = false;
+                    kvo.val_bool = false;
                 } else {
                     fprintf(stderr, "error: Invalid boolean value for KV override: %s\n", argv[i]);
                     invalid_param = true;

llm/filetype.go

@@ -0,0 +1,165 @@
+package llm
+
+import "fmt"
+
+type fileType uint32
+
+const (
+	fileTypeF32 fileType = iota
+	fileTypeF16
+	fileTypeQ4_0
+	fileTypeQ4_1
+	fileTypeQ4_1_F16
+	fileTypeQ4_2 // unused
+	fileTypeQ4_3 // unused
+	fileTypeQ8_0
+	fileTypeQ5_0
+	fileTypeQ5_1
+	fileTypeQ2_K
+	fileTypeQ3_K_S
+	fileTypeQ3_K_M
+	fileTypeQ3_K_L
+	fileTypeQ4_K_S
+	fileTypeQ4_K_M
+	fileTypeQ5_K_S
+	fileTypeQ5_K_M
+	fileTypeQ6_K
+	fileTypeIQ2_XXS
+	fileTypeIQ2_XS
+	fileTypeQ2_K_S
+	fileTypeQ3_K_XS
+	fileTypeIQ3_XXS
+	fileTypeIQ1_S
+	fileTypeIQ4_NL
+	fileTypeIQ3_S
+	fileTypeIQ2_S
+	fileTypeIQ4_XS
+
+	fileTypeUnknown
+)
+
+func ParseFileType(s string) (fileType, error) {
+	switch s {
+	case "F32":
+		return fileTypeF32, nil
+	case "F16":
+		return fileTypeF16, nil
+	case "Q4_0":
+		return fileTypeQ4_0, nil
+	case "Q4_1":
+		return fileTypeQ4_1, nil
+	case "Q4_1_F16":
+		return fileTypeQ4_1_F16, nil
+	case "Q8_0":
+		return fileTypeQ8_0, nil
+	case "Q5_0":
+		return fileTypeQ5_0, nil
+	case "Q5_1":
+		return fileTypeQ5_1, nil
+	case "Q2_K":
+		return fileTypeQ2_K, nil
+	case "Q3_K_S":
+		return fileTypeQ3_K_S, nil
+	case "Q3_K_M":
+		return fileTypeQ3_K_M, nil
+	case "Q3_K_L":
+		return fileTypeQ3_K_L, nil
+	case "Q4_K_S":
+		return fileTypeQ4_K_S, nil
+	case "Q4_K_M":
+		return fileTypeQ4_K_M, nil
+	case "Q5_K_S":
+		return fileTypeQ5_K_S, nil
+	case "Q5_K_M":
+		return fileTypeQ5_K_M, nil
+	case "Q6_K":
+		return fileTypeQ6_K, nil
+	case "IQ2_XXS":
+		return fileTypeIQ2_XXS, nil
+	case "IQ2_XS":
+		return fileTypeIQ2_XS, nil
+	case "Q2_K_S":
+		return fileTypeQ2_K_S, nil
+	case "Q3_K_XS":
+		return fileTypeQ3_K_XS, nil
+	case "IQ1_S":
+		return fileTypeIQ1_S, nil
+	case "IQ4_NL":
+		return fileTypeIQ4_NL, nil
+	case "IQ3_S":
+		return fileTypeIQ3_S, nil
+	case "IQ2_S":
+		return fileTypeIQ2_S, nil
+	case "IQ4_XS":
+		return fileTypeIQ4_XS, nil
+	case "IQ3_XXS":
+		return fileTypeIQ3_XXS, nil
+	default:
+		return fileTypeUnknown, fmt.Errorf("unknown fileType: %s", s)
+	}
+}
+
+func (t fileType) String() string {
+	switch t {
+	case fileTypeF32:
+		return "F32"
+	case fileTypeF16:
+		return "F16"
+	case fileTypeQ4_0:
+		return "Q4_0"
+	case fileTypeQ4_1:
+		return "Q4_1"
+	case fileTypeQ4_1_F16:
+		return "Q4_1_F16"
+	case fileTypeQ8_0:
+		return "Q8_0"
+	case fileTypeQ5_0:
+		return "Q5_0"
+	case fileTypeQ5_1:
+		return "Q5_1"
+	case fileTypeQ2_K:
+		return "Q2_K"
+	case fileTypeQ3_K_S:
+		return "Q3_K_S"
+	case fileTypeQ3_K_M:
+		return "Q3_K_M"
+	case fileTypeQ3_K_L:
+		return "Q3_K_L"
+	case fileTypeQ4_K_S:
+		return "Q4_K_S"
+	case fileTypeQ4_K_M:
+		return "Q4_K_M"
+	case fileTypeQ5_K_S:
+		return "Q5_K_S"
+	case fileTypeQ5_K_M:
+		return "Q5_K_M"
+	case fileTypeQ6_K:
+		return "Q6_K"
+	case fileTypeIQ2_XXS:
+		return "IQ2_XXS"
+	case fileTypeIQ2_XS:
+		return "IQ2_XS"
+	case fileTypeQ2_K_S:
+		return "Q2_K_S"
+	case fileTypeQ3_K_XS:
+		return "Q3_K_XS"
+	case fileTypeIQ1_S:
+		return "IQ1_S"
+	case fileTypeIQ4_NL:
+		return "IQ4_NL"
+	case fileTypeIQ3_S:
+		return "IQ3_S"
+	case fileTypeIQ2_S:
+		return "IQ2_S"
+	case fileTypeIQ4_XS:
+		return "IQ4_XS"
+	case fileTypeIQ3_XXS:
+		return "IQ3_XXS"
+	default:
+		return "unknown"
+	}
+}
+
+func (t fileType) Value() uint32 {
+	return uint32(t)
+}

llm/generate/gen_windows.ps1

@@ -42,7 +42,7 @@ function init_vars {
         "-DLLAMA_NATIVE=off"
         )
     $script:commonCpuDefs = @("-DCMAKE_POSITION_INDEPENDENT_CODE=on")
-    $script:ARCH = "amd64" # arm not yet supported.
+    $script:ARCH = $Env:PROCESSOR_ARCHITECTURE.ToLower()
     $script:DIST_BASE = "${script:SRC_DIR}\dist\windows-${script:ARCH}\ollama_runners"
     md "$script:DIST_BASE" -ea 0 > $null
     if ($env:CGO_CFLAGS -contains "-g") {
@@ -213,11 +213,11 @@ function build_static() {
     }
 }
 
-function build_cpu() {
+function build_cpu($gen_arch) {
     if ((-not "${env:OLLAMA_SKIP_CPU_GENERATE}" ) -and ((-not "${env:OLLAMA_CPU_TARGET}") -or ("${env:OLLAMA_CPU_TARGET}" -eq "cpu"))) {
         # remaining llama.cpp builds use MSVC 
         init_vars
-        $script:cmakeDefs = $script:commonCpuDefs + @("-A", "x64", "-DLLAMA_AVX=off", "-DLLAMA_AVX2=off", "-DLLAMA_AVX512=off", "-DLLAMA_FMA=off", "-DLLAMA_F16C=off") + $script:cmakeDefs
+        $script:cmakeDefs = $script:commonCpuDefs + @("-A", $gen_arch, "-DLLAMA_AVX=off", "-DLLAMA_AVX2=off", "-DLLAMA_AVX512=off", "-DLLAMA_FMA=off", "-DLLAMA_F16C=off") + $script:cmakeDefs
         $script:buildDir="../build/windows/${script:ARCH}/cpu"
         $script:distDir="$script:DIST_BASE\cpu"
         write-host "Building LCD CPU"
@@ -349,11 +349,15 @@ if ($($args.count) -eq 0) {
     git_module_setup
     apply_patches
     build_static
-    build_cpu
-    build_cpu_avx
-    build_cpu_avx2
-    build_cuda
-    build_rocm
+    if ($script:ARCH -eq "arm64") {
+        build_cpu("ARM64")
+    } else { # amd64
+        build_cpu("x64")
+        build_cpu_avx
+        build_cpu_avx2
+        build_cuda
+        build_rocm
+    }
 
     cleanup
     write-host "`ngo generate completed.  LLM runners: $(get-childitem -path $script:DIST_BASE)"

llm/ggml.go

@@ -13,82 +13,6 @@ type GGML struct {
 	model
 }
 
-const (
-	fileTypeF32 uint32 = iota
-	fileTypeF16
-	fileTypeQ4_0
-	fileTypeQ4_1
-	fileTypeQ4_1_F16
-	fileTypeQ8_0 uint32 = iota + 2
-	fileTypeQ5_0
-	fileTypeQ5_1
-	fileTypeQ2_K
-	fileTypeQ3_K_S
-	fileTypeQ3_K_M
-	fileTypeQ3_K_L
-	fileTypeQ4_K_S
-	fileTypeQ4_K_M
-	fileTypeQ5_K_S
-	fileTypeQ5_K_M
-	fileTypeQ6_K
-	fileTypeIQ2_XXS
-	fileTypeIQ2_XS
-	fileTypeQ2_K_S
-	fileTypeQ3_K_XS
-	fileTypeIQ3_XXS
-)
-
-func fileType(fileType uint32) string {
-	switch fileType {
-	case fileTypeF32:
-		return "F32"
-	case fileTypeF16:
-		return "F16"
-	case fileTypeQ4_0:
-		return "Q4_0"
-	case fileTypeQ4_1:
-		return "Q4_1"
-	case fileTypeQ4_1_F16:
-		return "Q4_1_F16"
-	case fileTypeQ8_0:
-		return "Q8_0"
-	case fileTypeQ5_0:
-		return "Q5_0"
-	case fileTypeQ5_1:
-		return "Q5_1"
-	case fileTypeQ2_K:
-		return "Q2_K"
-	case fileTypeQ3_K_S:
-		return "Q3_K_S"
-	case fileTypeQ3_K_M:
-		return "Q3_K_M"
-	case fileTypeQ3_K_L:
-		return "Q3_K_L"
-	case fileTypeQ4_K_S:
-		return "Q4_K_S"
-	case fileTypeQ4_K_M:
-		return "Q4_K_M"
-	case fileTypeQ5_K_S:
-		return "Q5_K_S"
-	case fileTypeQ5_K_M:
-		return "Q5_K_M"
-	case fileTypeQ6_K:
-		return "Q6_K"
-	case fileTypeIQ2_XXS:
-		return "IQ2_XXS"
-	case fileTypeIQ2_XS:
-		return "IQ2_XS"
-	case fileTypeQ2_K_S:
-		return "Q2_K_S"
-	case fileTypeQ3_K_XS:
-		return "Q3_K_XS"
-	case fileTypeIQ3_XXS:
-		return "IQ3_XXS"
-	default:
-		return "unknown"
-	}
-}
-
 type model interface {
 	KV() KV
 	Tensors() Tensors
@@ -123,7 +47,7 @@ func (kv KV) ParameterCount() uint64 {
 
 func (kv KV) FileType() string {
 	if u64 := kv.u64("general.file_type"); u64 > 0 {
-		return fileType(uint32(u64))
+		return fileType(uint32(u64)).String()
 	}
 
 	return "unknown"
@@ -201,9 +125,9 @@ type Tensor struct {
 
 func (t Tensor) blockSize() uint64 {
 	switch {
-	case t.Kind < 2:
+	case t.Kind < 2 || (t.Kind > 23 && t.Kind < 29):
 		return 1
-	case t.Kind < 10:
+	case t.Kind < 10 || t.Kind == 20:
 		return 32
 	default:
 		return 256
@@ -248,6 +172,16 @@ func (t Tensor) typeSize() uint64 {
 		return 2 + 2*blockSize/8 + blockSize/32
 	case 18: // IQ3_XXS
 		return 2 + 3*blockSize/8
+	case 19: // IQ1_S
+		return 2 + blockSize/8 + blockSize/16
+	case 20: // IQ4_NL
+		return 2 + blockSize/2
+	case 21: // IQ3_S
+		return 2 + 2*blockSize/8 + blockSize/8 + blockSize/32 + 4
+	case 22: // IQ2_S
+		return 2 + blockSize/4 + blockSize/16
+	case 23: // IQ4_XS
+		return 4 + blockSize/2 + blockSize/64
 	default:
 		return 0
 	}
@@ -286,6 +220,23 @@ const (
 
 var ErrUnsupportedFormat = errors.New("unsupported model format")
 
+func DetectGGMLType(b []byte) string {
+	switch binary.LittleEndian.Uint32(b[:4]) {
+	case FILE_MAGIC_GGML:
+		return "ggml"
+	case FILE_MAGIC_GGMF:
+		return "ggmf"
+	case FILE_MAGIC_GGJT:
+		return "ggjt"
+	case FILE_MAGIC_GGLA:
+		return "ggla"
+	case FILE_MAGIC_GGUF_LE, FILE_MAGIC_GGUF_BE:
+		return "gguf"
+	default:
+		return ""
+	}
+}
+
 func DecodeGGML(rs io.ReadSeeker) (*GGML, int64, error) {
 	var magic uint32
 	if err := binary.Read(rs, binary.LittleEndian, &magic); err != nil {

llm/llm.go

@@ -4,6 +4,7 @@ package llm
 // #cgo darwin,arm64 LDFLAGS: ${SRCDIR}/build/darwin/arm64_static/libllama.a -lstdc++
 // #cgo darwin,amd64 LDFLAGS: ${SRCDIR}/build/darwin/x86_64_static/libllama.a -lstdc++
 // #cgo windows,amd64 LDFLAGS: ${SRCDIR}/build/windows/amd64_static/libllama.a -static -lstdc++
+// #cgo windows,arm64 LDFLAGS: ${SRCDIR}/build/windows/arm64_static/libllama.a -static -lstdc++
 // #cgo linux,amd64 LDFLAGS: ${SRCDIR}/build/linux/x86_64_static/libllama.a -lstdc++
 // #cgo linux,arm64 LDFLAGS: ${SRCDIR}/build/linux/arm64_static/libllama.a -lstdc++
 // #include <stdlib.h>
@@ -19,7 +20,7 @@ func SystemInfo() string {
 	return C.GoString(C.llama_print_system_info())
 }
 
-func Quantize(infile, outfile, filetype string) error {
+func Quantize(infile, outfile string, ftype fileType) error {
 	cinfile := C.CString(infile)
 	defer C.free(unsafe.Pointer(cinfile))
 
@@ -28,58 +29,10 @@ func Quantize(infile, outfile, filetype string) error {
 
 	params := C.llama_model_quantize_default_params()
 	params.nthread = -1
+	params.ftype = ftype.Value()
 
-	switch filetype {
-	case "F32":
-		params.ftype = fileTypeF32
-	case "F16":
-		params.ftype = fileTypeF16
-	case "Q4_0":
-		params.ftype = fileTypeQ4_0
-	case "Q4_1":
-		params.ftype = fileTypeQ4_1
-	case "Q4_1_F16":
-		params.ftype = fileTypeQ4_1_F16
-	case "Q8_0":
-		params.ftype = fileTypeQ8_0
-	case "Q5_0":
-		params.ftype = fileTypeQ5_0
-	case "Q5_1":
-		params.ftype = fileTypeQ5_1
-	case "Q2_K":
-		params.ftype = fileTypeQ2_K
-	case "Q3_K_S":
-		params.ftype = fileTypeQ3_K_S
-	case "Q3_K_M":
-		params.ftype = fileTypeQ3_K_M
-	case "Q3_K_L":
-		params.ftype = fileTypeQ3_K_L
-	case "Q4_K_S":
-		params.ftype = fileTypeQ4_K_S
-	case "Q4_K_M":
-		params.ftype = fileTypeQ4_K_M
-	case "Q5_K_S":
-		params.ftype = fileTypeQ5_K_S
-	case "Q5_K_M":
-		params.ftype = fileTypeQ5_K_M
-	case "Q6_K":
-		params.ftype = fileTypeQ6_K
-	case "IQ2_XXS":
-		params.ftype = fileTypeIQ2_XXS
-	case "IQ2_XS":
-		params.ftype = fileTypeIQ2_XS
-	case "Q2_K_S":
-		params.ftype = fileTypeQ2_K_S
-	case "Q3_K_XS":
-		params.ftype = fileTypeQ3_K_XS
-	case "IQ3_XXS":
-		params.ftype = fileTypeIQ3_XXS
-	default:
-		return fmt.Errorf("unknown filetype: %s", filetype)
-	}
-
-	if retval := C.llama_model_quantize(cinfile, coutfile, &params); retval != 0 {
-		return fmt.Errorf("llama_model_quantize: %d", retval)
+	if rc := C.llama_model_quantize(cinfile, coutfile, &params); rc != 0 {
+		return fmt.Errorf("llama_model_quantize: %d", rc)
 	}
 
 	return nil

llm/memory.go

@@ -88,6 +88,11 @@ func EstimateGPULayers(gpus []gpu.GpuInfo, ggml *GGML, projectors []string, opts
 	graphFullOffload *= uint64(len(gpus))
 	graphPartialOffload *= uint64(len(gpus))
 
+	// on metal there's no partial offload overhead
+	if gpus[0].Library == "metal" {
+		graphPartialOffload = graphFullOffload
+	}
+
 	// memoryRequiredTotal represents the memory required for full GPU offloading (all layers)
 	memoryRequiredTotal := memoryMinimum + graphFullOffload
 

llm/server.go

@@ -73,8 +73,7 @@ func LoadModel(model string) (*GGML, error) {
 func NewLlamaServer(gpus gpu.GpuInfoList, model string, ggml *GGML, adapters, projectors []string, opts api.Options) (LlamaServer, error) {
 	var err error
 	if opts.NumCtx > int(ggml.KV().ContextLength()) {
-		slog.Warn("requested context length is greater than model max context length", "requested", opts.NumCtx, "model", ggml.KV().ContextLength())
-		opts.NumCtx = int(ggml.KV().ContextLength())
+		slog.Warn("requested context length is greater than the model's training context window size", "requested", opts.NumCtx, "training size", ggml.KV().ContextLength())
 	}
 
 	if opts.NumCtx < 4 {
@@ -301,12 +300,6 @@ func NewLlamaServer(gpus gpu.GpuInfoList, model string, ggml *GGML, adapters, pr
 			continue
 		}
 
-		// reap subprocess when it exits
-		go func() {
-			// Exit status managed via getServerStatus
-			_ = s.cmd.Wait()
-		}()
-
 		// TODO - make sure this is all wired up correctly
 		// if err = s.WaitUntilRunning(); err != nil {
 		// 	slog.Error("error starting llama server", "server", servers[i], "error", err)
@@ -900,7 +893,13 @@ func (s *llmServer) Detokenize(ctx context.Context, tokens []int) (string, error
 func (s *llmServer) Close() error {
 	if s.cmd != nil {
 		slog.Debug("stopping llama server")
-		return s.cmd.Process.Kill()
+		if err := s.cmd.Process.Kill(); err != nil {
+			return err
+		}
+
+		_ = s.cmd.Wait()
+
+		slog.Debug("llama server stopped")
 	}
 
 	return nil

macapp/.eslintrc.json

@@ -0,0 +1,16 @@
+{
+  "env": {
+    "browser": true,
+    "es6": true,
+    "node": true
+  },
+  "extends": [
+    "eslint:recommended",
+    "plugin:@typescript-eslint/eslint-recommended",
+    "plugin:@typescript-eslint/recommended",
+    "plugin:import/recommended",
+    "plugin:import/electron",
+    "plugin:import/typescript"
+  ],
+  "parser": "@typescript-eslint/parser"
+}

macapp/.gitignore

@@ -0,0 +1,92 @@
+# Logs
+logs
+*.log
+npm-debug.log*
+yarn-debug.log*
+yarn-error.log*
+lerna-debug.log*
+
+# Diagnostic reports (https://nodejs.org/api/report.html)
+report.[0-9]*.[0-9]*.[0-9]*.[0-9]*.json
+
+# Runtime data
+pids
+*.pid
+*.seed
+*.pid.lock
+.DS_Store
+
+# Directory for instrumented libs generated by jscoverage/JSCover
+lib-cov
+
+# Coverage directory used by tools like istanbul
+coverage
+*.lcov
+
+# nyc test coverage
+.nyc_output
+
+# node-waf configuration
+.lock-wscript
+
+# Compiled binary addons (https://nodejs.org/api/addons.html)
+build/Release
+
+# Dependency directories
+node_modules/
+jspm_packages/
+
+# TypeScript v1 declaration files
+typings/
+
+# TypeScript cache
+*.tsbuildinfo
+
+# Optional npm cache directory
+.npm
+
+# Optional eslint cache
+.eslintcache
+
+# Optional REPL history
+.node_repl_history
+
+# Output of 'npm pack'
+*.tgz
+
+# Yarn Integrity file
+.yarn-integrity
+
+# dotenv environment variables file
+.env
+.env.test
+
+# parcel-bundler cache (https://parceljs.org/)
+.cache
+
+# next.js build output
+.next
+
+# nuxt.js build output
+.nuxt
+
+# vuepress build output
+.vuepress/dist
+
+# Serverless directories
+.serverless/
+
+# FuseBox cache
+.fusebox/
+
+# DynamoDB Local files
+.dynamodb/
+
+# Webpack
+.webpack/
+
+# Vite
+.vite/
+
+# Electron-Forge
+out/

macapp/README.md

@@ -0,0 +1,21 @@
+# Desktop
+
+This app builds upon Ollama to provide a desktop experience for running models.
+
+## Developing
+
+First, build the `ollama` binary:
+
+```
+cd ..
+go build .
+```
+
+Then run the desktop app with `npm start`:
+
+```
+cd macapp
+npm install
+npm start
+```
+

macapp/forge.config.ts

@@ -0,0 +1,78 @@
+import type { ForgeConfig } from '@electron-forge/shared-types'
+import { MakerSquirrel } from '@electron-forge/maker-squirrel'
+import { MakerZIP } from '@electron-forge/maker-zip'
+import { PublisherGithub } from '@electron-forge/publisher-github'
+import { AutoUnpackNativesPlugin } from '@electron-forge/plugin-auto-unpack-natives'
+import { WebpackPlugin } from '@electron-forge/plugin-webpack'
+import * as path from 'path'
+import * as fs from 'fs'
+
+import { mainConfig } from './webpack.main.config'
+import { rendererConfig } from './webpack.renderer.config'
+
+const packageJson = JSON.parse(fs.readFileSync(path.resolve(__dirname, './package.json'), 'utf8'))
+
+const config: ForgeConfig = {
+  packagerConfig: {
+    appVersion: process.env.VERSION || packageJson.version,
+    asar: true,
+    icon: './assets/icon.icns',
+    extraResource: [
+      '../dist/ollama',
+      path.join(__dirname, './assets/iconTemplate.png'),
+      path.join(__dirname, './assets/iconTemplate@2x.png'),
+      path.join(__dirname, './assets/iconUpdateTemplate.png'),
+      path.join(__dirname, './assets/iconUpdateTemplate@2x.png'),
+      path.join(__dirname, './assets/iconDarkTemplate.png'),
+      path.join(__dirname, './assets/iconDarkTemplate@2x.png'),
+      path.join(__dirname, './assets/iconDarkUpdateTemplate.png'),
+      path.join(__dirname, './assets/iconDarkUpdateTemplate@2x.png'),
+    ],
+    ...(process.env.SIGN
+      ? {
+          osxSign: {
+            identity: process.env.APPLE_IDENTITY,
+          },
+          osxNotarize: {
+            tool: 'notarytool',
+            appleId: process.env.APPLE_ID || '',
+            appleIdPassword: process.env.APPLE_PASSWORD || '',
+            teamId: process.env.APPLE_TEAM_ID || '',
+          },
+        }
+      : {}),
+    osxUniversal: {
+      x64ArchFiles: '**/ollama',
+    },
+  },
+  rebuildConfig: {},
+  makers: [new MakerSquirrel({}), new MakerZIP({}, ['darwin'])],
+  hooks: {
+    readPackageJson: async (_, packageJson) => {
+      return { ...packageJson, version: process.env.VERSION || packageJson.version }
+    },
+  },
+  plugins: [
+    new AutoUnpackNativesPlugin({}),
+    new WebpackPlugin({
+      mainConfig,
+      devContentSecurityPolicy: `default-src * 'unsafe-eval' 'unsafe-inline'; img-src data: 'self'`,
+      renderer: {
+        config: rendererConfig,
+        nodeIntegration: true,
+        entryPoints: [
+          {
+            html: './src/index.html',
+            js: './src/renderer.tsx',
+            name: 'main_window',
+            preload: {
+              js: './src/preload.ts',
+            },
+          },
+        ],
+      },
+    }),
+  ],
+}
+
+export default config

macapp/package.json

@@ -0,0 +1,84 @@
+{
+  "name": "ollama",
+  "productName": "Ollama",
+  "version": "0.0.0",
+  "description": "ollama",
+  "main": ".webpack/main",
+  "scripts": {
+    "start": "electron-forge start",
+    "package": "electron-forge package --arch universal",
+    "package:sign": "SIGN=1 electron-forge package --arch universal",
+    "make": "electron-forge make --arch universal",
+    "make:sign": "SIGN=1 electron-forge make --arch universal",
+    "publish": "SIGN=1 electron-forge publish",
+    "lint": "eslint --ext .ts,.tsx .",
+    "format": "prettier --check . --ignore-path .gitignore",
+    "format:fix": "prettier --write . --ignore-path .gitignore"
+  },
+  "keywords": [],
+  "author": {
+    "name": "Jeffrey Morgan",
+    "email": "jmorganca@gmail.com"
+  },
+  "license": "MIT",
+  "devDependencies": {
+    "@babel/core": "^7.22.5",
+    "@babel/preset-react": "^7.22.5",
+    "@electron-forge/cli": "^6.2.1",
+    "@electron-forge/maker-deb": "^6.2.1",
+    "@electron-forge/maker-rpm": "^6.2.1",
+    "@electron-forge/maker-squirrel": "^6.2.1",
+    "@electron-forge/maker-zip": "^6.2.1",
+    "@electron-forge/plugin-auto-unpack-natives": "^6.2.1",
+    "@electron-forge/plugin-webpack": "^6.2.1",
+    "@electron-forge/publisher-github": "^6.2.1",
+    "@electron/universal": "^1.4.1",
+    "@svgr/webpack": "^8.0.1",
+    "@types/chmodr": "^1.0.0",
+    "@types/node": "^20.4.0",
+    "@types/react": "^18.2.14",
+    "@types/react-dom": "^18.2.6",
+    "@types/uuid": "^9.0.2",
+    "@typescript-eslint/eslint-plugin": "^5.60.0",
+    "@typescript-eslint/parser": "^5.60.0",
+    "@vercel/webpack-asset-relocator-loader": "^1.7.3",
+    "babel-loader": "^9.1.2",
+    "chmodr": "^1.2.0",
+    "copy-webpack-plugin": "^11.0.0",
+    "css-loader": "^6.8.1",
+    "electron": "25.9.2",
+    "eslint": "^8.43.0",
+    "eslint-plugin-import": "^2.27.5",
+    "fork-ts-checker-webpack-plugin": "^7.3.0",
+    "node-loader": "^2.0.0",
+    "postcss": "^8.4.24",
+    "postcss-import": "^15.1.0",
+    "postcss-loader": "^7.3.3",
+    "postcss-preset-env": "^8.5.1",
+    "prettier": "^2.8.8",
+    "prettier-plugin-tailwindcss": "^0.3.0",
+    "style-loader": "^3.3.3",
+    "svg-inline-loader": "^0.8.2",
+    "tailwindcss": "^3.3.2",
+    "ts-loader": "^9.4.3",
+    "ts-node": "^10.9.1",
+    "typescript": "~4.5.4",
+    "url-loader": "^4.1.1",
+    "webpack": "^5.88.0",
+    "webpack-cli": "^5.1.4",
+    "webpack-dev-server": "^4.15.1"
+  },
+  "dependencies": {
+    "@electron/remote": "^2.0.10",
+    "@heroicons/react": "^2.0.18",
+    "@segment/analytics-node": "^1.0.0",
+    "copy-to-clipboard": "^3.3.3",
+    "electron-squirrel-startup": "^1.0.0",
+    "electron-store": "^8.1.0",
+    "react": "^18.2.0",
+    "react-dom": "^18.2.0",
+    "uuid": "^9.0.0",
+    "winston": "^3.10.0",
+    "winston-daily-rotate-file": "^4.7.1"
+  }
+}

macapp/postcss.config.js

@@ -0,0 +1,7 @@
+module.exports = {
+  plugins: {
+    'postcss-import': {},
+    tailwindcss: {},
+    autoprefixer: {},
+  },
+}

macapp/src/app.css

@@ -0,0 +1,34 @@
+@tailwind base;
+@tailwind components;
+@tailwind utilities;
+
+html,
+body {
+  background: transparent;
+}
+
+.drag {
+  -webkit-app-region: drag;
+}
+
+.no-drag {
+  -webkit-app-region: no-drag;
+}
+
+.blink {
+  -webkit-animation: 1s blink step-end infinite;
+  -moz-animation: 1s blink step-end infinite;
+  -ms-animation: 1s blink step-end infinite;
+  -o-animation: 1s blink step-end infinite;
+  animation: 1s blink step-end infinite;
+}
+
+@keyframes blink {
+  from,
+  to {
+    color: transparent;
+  }
+  50% {
+    color: black;
+  }
+}

macapp/src/app.tsx

@@ -0,0 +1,122 @@
+import { useState } from 'react'
+import copy from 'copy-to-clipboard'
+import { CheckIcon, DocumentDuplicateIcon } from '@heroicons/react/24/outline'
+import Store from 'electron-store'
+import { getCurrentWindow, app } from '@electron/remote'
+
+import { install } from './install'
+import OllamaIcon from './ollama.svg'
+
+const store = new Store()
+
+enum Step {
+  WELCOME = 0,
+  CLI,
+  FINISH,
+}
+
+export default function () {
+  const [step, setStep] = useState<Step>(Step.WELCOME)
+  const [commandCopied, setCommandCopied] = useState<boolean>(false)
+
+  const command = 'ollama run llama3'
+
+  return (
+    <div className='drag'>
+      <div className='mx-auto flex min-h-screen w-full flex-col justify-between bg-white px-4 pt-16'>
+        {step === Step.WELCOME && (
+          <>
+            <div className='mx-auto text-center'>
+              <h1 className='mb-6 mt-4 text-2xl tracking-tight text-gray-900'>Welcome to Ollama</h1>
+              <p className='mx-auto w-[65%] text-sm text-gray-400'>
+                Let's get you up and running with your own large language models.
+              </p>
+              <button
+                onClick={() => setStep(Step.CLI)}
+                className='no-drag rounded-dm mx-auto my-8 w-[40%] rounded-md bg-black px-4 py-2 text-sm text-white hover:brightness-110'
+              >
+                Next
+              </button>
+            </div>
+            <div className='mx-auto'>
+              <OllamaIcon />
+            </div>
+          </>
+        )}
+        {step === Step.CLI && (
+          <>
+            <div className='mx-auto flex flex-col space-y-28 text-center'>
+              <h1 className='mt-4 text-2xl tracking-tight text-gray-900'>Install the command line</h1>
+              <pre className='mx-auto text-4xl text-gray-400'>&gt; ollama</pre>
+              <div className='mx-auto'>
+                <button
+                  onClick={async () => {
+                    try {
+                      await install()
+                      setStep(Step.FINISH)
+                    } catch (e) {
+                      console.error('could not install: ', e)
+                    } finally {
+                      getCurrentWindow().show()
+                      getCurrentWindow().focus()
+                    }
+                  }}
+                  className='no-drag rounded-dm mx-auto w-[60%] rounded-md bg-black px-4 py-2 text-sm text-white hover:brightness-110'
+                >
+                  Install
+                </button>
+                <p className='mx-auto my-4 w-[70%] text-xs text-gray-400'>
+                  You will be prompted for administrator access
+                </p>
+              </div>
+            </div>
+          </>
+        )}
+        {step === Step.FINISH && (
+          <>
+            <div className='mx-auto flex flex-col space-y-20 text-center'>
+              <h1 className='mt-4 text-2xl tracking-tight text-gray-900'>Run your first model</h1>
+              <div className='flex flex-col'>
+                <div className='group relative flex items-center'>
+                  <pre className='language-none text-2xs w-full rounded-md bg-gray-100 px-4 py-3 text-start leading-normal'>
+                    {command}
+                  </pre>
+                  <button
+                    className={`no-drag absolute right-[5px] px-2 py-2 ${
+                      commandCopied
+                        ? 'text-gray-900 opacity-100 hover:cursor-auto'
+                        : 'text-gray-200 opacity-50 hover:cursor-pointer'
+                    } hover:font-bold hover:text-gray-900 group-hover:opacity-100`}
+                    onClick={() => {
+                      copy(command)
+                      setCommandCopied(true)
+                      setTimeout(() => setCommandCopied(false), 3000)
+                    }}
+                  >
+                    {commandCopied ? (
+                      <CheckIcon className='h-4 w-4 font-bold text-gray-500' />
+                    ) : (
+                      <DocumentDuplicateIcon className='h-4 w-4 text-gray-500' />
+                    )}
+                  </button>
+                </div>
+                <p className='mx-auto my-4 w-[70%] text-xs text-gray-400'>
+                  Run this command in your favorite terminal.
+                </p>
+              </div>
+              <button
+                onClick={() => {
+                  store.set('first-time-run', true)
+                  window.close()
+                }}
+                className='no-drag rounded-dm mx-auto w-[60%] rounded-md bg-black px-4 py-2 text-sm text-white hover:brightness-110'
+              >
+                Finish
+              </button>
+            </div>
+          </>
+        )}
+      </div>
+    </div>
+  )
+}

macapp/src/declarations.d.ts

@@ -0,0 +1,4 @@
+declare module '*.svg' {
+  const content: string
+  export default content
+}

macapp/src/index.html

@@ -0,0 +1,9 @@
+<!DOCTYPE html>
+<html>
+  <head>
+    <meta charset="UTF-8" />
+  </head>
+  <body>
+    <div id="app"></div>
+  </body>
+</html>

macapp/src/index.ts

@@ -0,0 +1,302 @@
+import { spawn, ChildProcess } from 'child_process'
+import { app, autoUpdater, dialog, Tray, Menu, BrowserWindow, MenuItemConstructorOptions, nativeTheme } from 'electron'
+import Store from 'electron-store'
+import winston from 'winston'
+import 'winston-daily-rotate-file'
+import * as path from 'path'
+
+import { v4 as uuidv4 } from 'uuid'
+import { installed } from './install'
+
+require('@electron/remote/main').initialize()
+
+if (require('electron-squirrel-startup')) {
+  app.quit()
+}
+
+const store = new Store()
+
+let welcomeWindow: BrowserWindow | null = null
+
+declare const MAIN_WINDOW_WEBPACK_ENTRY: string
+
+const logger = winston.createLogger({
+  transports: [
+    new winston.transports.Console(),
+    new winston.transports.File({
+      filename: path.join(app.getPath('home'), '.ollama', 'logs', 'server.log'),
+      maxsize: 1024 * 1024 * 20,
+      maxFiles: 5,
+    }),
+  ],
+  format: winston.format.printf(info => info.message),
+})
+
+app.on('ready', () => {
+  const gotTheLock = app.requestSingleInstanceLock()
+  if (!gotTheLock) {
+    app.exit(0)
+    return
+  }
+
+  app.on('second-instance', () => {
+    if (app.hasSingleInstanceLock()) {
+      app.releaseSingleInstanceLock()
+    }
+
+    if (proc) {
+      proc.off('exit', restart)
+      proc.kill()
+    }
+
+    app.exit(0)
+  })
+
+  app.focus({ steal: true })
+
+  init()
+})
+
+function firstRunWindow() {
+  // Create the browser window.
+  welcomeWindow = new BrowserWindow({
+    width: 400,
+    height: 500,
+    frame: false,
+    fullscreenable: false,
+    resizable: false,
+    movable: true,
+    show: false,
+    webPreferences: {
+      nodeIntegration: true,
+      contextIsolation: false,
+    },
+  })
+
+  require('@electron/remote/main').enable(welcomeWindow.webContents)
+
+  welcomeWindow.loadURL(MAIN_WINDOW_WEBPACK_ENTRY)
+  welcomeWindow.on('ready-to-show', () => welcomeWindow.show())
+  welcomeWindow.on('closed', () => {
+    if (process.platform === 'darwin') {
+      app.dock.hide()
+    }
+  })
+}
+
+let tray: Tray | null = null
+let updateAvailable = false
+const assetPath = app.isPackaged ? process.resourcesPath : path.join(__dirname, '..', '..', 'assets')
+
+function trayIconPath() {
+  return nativeTheme.shouldUseDarkColors
+    ? updateAvailable
+      ? path.join(assetPath, 'iconDarkUpdateTemplate.png')
+      : path.join(assetPath, 'iconDarkTemplate.png')
+    : updateAvailable
+    ? path.join(assetPath, 'iconUpdateTemplate.png')
+    : path.join(assetPath, 'iconTemplate.png')
+}
+
+function updateTrayIcon() {
+  if (tray) {
+    tray.setImage(trayIconPath())
+  }
+}
+
+function updateTray() {
+  const updateItems: MenuItemConstructorOptions[] = [
+    { label: 'An update is available', enabled: false },
+    {
+      label: 'Restart to update',
+      click: () => autoUpdater.quitAndInstall(),
+    },
+    { type: 'separator' },
+  ]
+
+  const menu = Menu.buildFromTemplate([
+    ...(updateAvailable ? updateItems : []),
+    { role: 'quit', label: 'Quit Ollama', accelerator: 'Command+Q' },
+  ])
+
+  if (!tray) {
+    tray = new Tray(trayIconPath())
+  }
+
+  tray.setToolTip(updateAvailable ? 'An update is available' : 'Ollama')
+  tray.setContextMenu(menu)
+  tray.setImage(trayIconPath())
+
+  nativeTheme.off('updated', updateTrayIcon)
+  nativeTheme.on('updated', updateTrayIcon)
+}
+
+let proc: ChildProcess = null
+
+function server() {
+  const binary = app.isPackaged
+    ? path.join(process.resourcesPath, 'ollama')
+    : path.resolve(process.cwd(), '..', 'ollama')
+
+  proc = spawn(binary, ['serve'])
+
+  proc.stdout.on('data', data => {
+    logger.info(data.toString().trim())
+  })
+
+  proc.stderr.on('data', data => {
+    logger.error(data.toString().trim())
+  })
+
+  proc.on('exit', restart)
+}
+
+function restart() {
+  setTimeout(server, 1000)
+}
+
+app.on('before-quit', () => {
+  if (proc) {
+    proc.off('exit', restart)
+    proc.kill('SIGINT') // send SIGINT signal to the server, which also stops any loaded llms
+  }
+})
+
+const updateURL = `https://ollama.ai/api/update?os=${process.platform}&arch=${
+  process.arch
+}&version=${app.getVersion()}&id=${id()}`
+
+let latest = ''
+async function isNewReleaseAvailable() {
+  try {
+    const response = await fetch(updateURL)
+
+    if (!response.ok) {
+      return false
+    }
+
+    if (response.status === 204) {
+      return false
+    }
+
+    const data = await response.json()
+
+    const url = data?.url
+    if (!url) {
+      return false
+    }
+
+    if (latest === url) {
+      return false
+    }
+
+    latest = url
+
+    return true
+  } catch (error) {
+    logger.error(`update check failed - ${error}`)
+    return false
+  }
+}
+
+async function checkUpdate() {
+  const available = await isNewReleaseAvailable()
+  if (available) {
+    logger.info('checking for update')
+    autoUpdater.checkForUpdates()
+  }
+}
+
+function init() {
+  if (app.isPackaged) {
+    checkUpdate()
+    setInterval(() => {
+      checkUpdate()
+    }, 60 * 60 * 1000)
+  }
+
+  updateTray()
+
+  if (process.platform === 'darwin') {
+    if (app.isPackaged) {
+      if (!app.isInApplicationsFolder()) {
+        const chosen = dialog.showMessageBoxSync({
+          type: 'question',
+          buttons: ['Move to Applications', 'Do Not Move'],
+          message: 'Ollama works best when run from the Applications directory.',
+          defaultId: 0,
+          cancelId: 1,
+        })
+
+        if (chosen === 0) {
+          try {
+            app.moveToApplicationsFolder({
+              conflictHandler: conflictType => {
+                if (conflictType === 'existsAndRunning') {
+                  dialog.showMessageBoxSync({
+                    type: 'info',
+                    message: 'Cannot move to Applications directory',
+                    detail:
+                      'Another version of Ollama is currently running from your Applications directory. Close it first and try again.',
+                  })
+                }
+                return true
+              },
+            })
+            return
+          } catch (e) {
+            logger.error(`[Move to Applications] Failed to move to applications folder - ${e.message}}`)
+          }
+        }
+      }
+    }
+  }
+
+  server()
+
+  if (store.get('first-time-run') && installed()) {
+    if (process.platform === 'darwin') {
+      app.dock.hide()
+    }
+
+    app.setLoginItemSettings({ openAtLogin: app.getLoginItemSettings().openAtLogin })
+    return
+  }
+
+  // This is the first run or the CLI is no longer installed
+  app.setLoginItemSettings({ openAtLogin: true })
+  firstRunWindow()
+}
+
+// Quit when all windows are closed, except on macOS. There, it's common
+// for applications and their menu bar to stay active until the user quits
+// explicitly with Cmd + Q.
+app.on('window-all-closed', () => {
+  if (process.platform !== 'darwin') {
+    app.quit()
+  }
+})
+
+function id(): string {
+  const id = store.get('id') as string
+
+  if (id) {
+    return id
+  }
+
+  const uuid = uuidv4()
+  store.set('id', uuid)
+  return uuid
+}
+
+autoUpdater.setFeedURL({ url: updateURL })
+
+autoUpdater.on('error', e => {
+  logger.error(`update check failed - ${e.message}`)
+  console.error(`update check failed - ${e.message}`)
+})
+
+autoUpdater.on('update-downloaded', () => {
+  updateAvailable = true
+  updateTray()
+})

macapp/src/install.ts

@@ -0,0 +1,21 @@
+import * as fs from 'fs'
+import { exec as cbExec } from 'child_process'
+import * as path from 'path'
+import { promisify } from 'util'
+
+const app = process && process.type === 'renderer' ? require('@electron/remote').app : require('electron').app
+const ollama = app.isPackaged ? path.join(process.resourcesPath, 'ollama') : path.resolve(process.cwd(), '..', 'ollama')
+const exec = promisify(cbExec)
+const symlinkPath = '/usr/local/bin/ollama'
+
+export function installed() {
+  return fs.existsSync(symlinkPath) && fs.readlinkSync(symlinkPath) === ollama
+}
+
+export async function install() {
+  const command = `do shell script "mkdir -p ${path.dirname(
+    symlinkPath
+  )} && ln -F -s \\"${ollama}\\" \\"${symlinkPath}\\"" with administrator privileges`
+
+  await exec(`osascript -e '${command}'`)
+}