cmd.go 44.3 KB
Newer Older
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1
2
3
package cmd

import (
Michael Yang's avatar
Michael Yang committed
4
	"bufio"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
5
	"context"
6
7
	"crypto/ed25519"
	"crypto/rand"
8
	"encoding/base64"
9
	"encoding/json"
10
	"encoding/pem"
Michael Yang's avatar
Michael Yang committed
11
	"errors"
Bruce MacDonald's avatar
Bruce MacDonald committed
12
	"fmt"
Michael Yang's avatar
Michael Yang committed
13
	"io"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
14
	"log"
15
	"math"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
16
	"net"
17
	"net/http"
18
	"net/url"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
19
	"os"
20
	"os/signal"
21
	"path/filepath"
22
	"runtime"
23
	"slices"
24
	"sort"
Michael Yang's avatar
Michael Yang committed
25
	"strconv"
Michael Yang's avatar
Michael Yang committed
26
	"strings"
27
	"sync/atomic"
28
	"syscall"
Michael Yang's avatar
Michael Yang committed
29
	"time"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
30

31
	"github.com/containerd/console"
32
	"github.com/mattn/go-runewidth"
Patrick Devine's avatar
Patrick Devine committed
33
	"github.com/olekukonko/tablewriter"
Michael Yang's avatar
Michael Yang committed
34
	"github.com/spf13/cobra"
35
	"golang.org/x/crypto/ssh"
36
	"golang.org/x/sync/errgroup"
37
	"golang.org/x/term"
Michael Yang's avatar
Michael Yang committed
38

39
	"github.com/ollama/ollama/api"
40
	"github.com/ollama/ollama/auth"
41
	"github.com/ollama/ollama/envconfig"
42
	"github.com/ollama/ollama/format"
43
	"github.com/ollama/ollama/parser"
44
	"github.com/ollama/ollama/progress"
45
	"github.com/ollama/ollama/readline"
Jesse Gross's avatar
Jesse Gross committed
46
	"github.com/ollama/ollama/runner"
47
	"github.com/ollama/ollama/server"
48
	"github.com/ollama/ollama/types/model"
49
	"github.com/ollama/ollama/types/syncmap"
50
	"github.com/ollama/ollama/version"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
51
52
)

53
54
const ConnectInstructions = "To sign in, navigate to:\n    https://ollama.com/connect?name=%s&key=%s\n\n"

55
56
57
58
59
60
61
62
63
// ensureThinkingSupport emits a warning if the model does not advertise thinking support
func ensureThinkingSupport(ctx context.Context, client *api.Client, name string) {
	if name == "" {
		return
	}
	resp, err := client.Show(ctx, &api.ShowRequest{Model: name})
	if err != nil {
		return
	}
64
65
	if slices.Contains(resp.Capabilities, model.CapabilityThinking) {
		return
66
67
68
69
	}
	fmt.Fprintf(os.Stderr, "warning: model %q does not support thinking output\n", name)
}

70
var errModelfileNotFound = errors.New("specified Modelfile wasn't found")
71
72

func getModelfileName(cmd *cobra.Command) (string, error) {
73
	filename, _ := cmd.Flags().GetString("file")
74
75
76
77
78
79

	if filename == "" {
		filename = "Modelfile"
	}

	absName, err := filepath.Abs(filename)
80
	if err != nil {
81
		return "", err
82
83
	}

84
	_, err = os.Stat(absName)
85
	if err != nil {
86
		return "", err
87
	}
88

89
90
91
92
	return absName, nil
}

func CreateHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
93
94
95
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

96
97
98
99
100
101
102
103
104
105
	var reader io.Reader

	filename, err := getModelfileName(cmd)
	if os.IsNotExist(err) {
		if filename == "" {
			reader = strings.NewReader("FROM .\n")
		} else {
			return errModelfileNotFound
		}
	} else if err != nil {
Michael Yang's avatar
Michael Yang committed
106
		return err
107
108
109
110
111
112
113
114
	} else {
		f, err := os.Open(filename)
		if err != nil {
			return err
		}

		reader = f
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
115
116
	}

117
	modelfile, err := parser.ParseFile(reader)
Michael Yang's avatar
Michael Yang committed
118
119
120
121
	if err != nil {
		return err
	}

122
123
124
125
	status := "gathering model components"
	spinner := progress.NewSpinner(status)
	p.Add(status, spinner)

126
	req, err := modelfile.CreateRequest(filepath.Dir(filename))
Michael Yang's avatar
Michael Yang committed
127
128
129
	if err != nil {
		return err
	}
130
	spinner.Stop()
Michael Yang's avatar
Michael Yang committed
131

132
	req.Model = args[0]
133
134
135
136
	quantize, _ := cmd.Flags().GetString("quantize")
	if quantize != "" {
		req.Quantize = quantize
	}
137

138
139
140
141
142
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

143
144
145
146
147
148
	var g errgroup.Group
	g.SetLimit(max(runtime.GOMAXPROCS(0)-1, 1))

	files := syncmap.NewSyncMap[string, string]()
	for f, digest := range req.Files {
		g.Go(func() error {
149
			if _, err := createBlob(cmd, client, f, digest, p); err != nil {
Michael Yang's avatar
Michael Yang committed
150
151
				return err
			}
152
153
154
155
156
157
158

			// TODO: this is incorrect since the file might be in a subdirectory
			//       instead this should take the path relative to the model directory
			//       but the current implementation does not allow this
			files.Store(filepath.Base(f), digest)
			return nil
		})
159
	}
Michael Yang's avatar
Michael Yang committed
160

161
162
163
	adapters := syncmap.NewSyncMap[string, string]()
	for f, digest := range req.Adapters {
		g.Go(func() error {
164
			if _, err := createBlob(cmd, client, f, digest, p); err != nil {
Michael Yang's avatar
Michael Yang committed
165
166
				return err
			}
167
168
169
170
171

			// TODO: same here
			adapters.Store(filepath.Base(f), digest)
			return nil
		})
Michael Yang's avatar
Michael Yang committed
172
	}
Michael Yang's avatar
Michael Yang committed
173

174
175
176
177
178
179
180
	if err := g.Wait(); err != nil {
		return err
	}

	req.Files = files.Items()
	req.Adapters = adapters.Items()

Michael Yang's avatar
Michael Yang committed
181
	bars := make(map[string]*progress.Bar)
182
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
183
184
185
		if resp.Digest != "" {
			bar, ok := bars[resp.Digest]
			if !ok {
186
187
188
189
190
				msg := resp.Status
				if msg == "" {
					msg = fmt.Sprintf("pulling %s...", resp.Digest[7:19])
				}
				bar = progress.NewBar(msg, resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
191
192
193
194
195
196
197
198
199
200
201
202
203
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
			spinner.Stop()

			status = resp.Status
			spinner = progress.NewSpinner(status)
			p.Add(status, spinner)
		}

204
205
206
		return nil
	}

207
	if err := client.Create(cmd.Context(), req, fn); err != nil {
208
209
210
		if strings.Contains(err.Error(), "path or Modelfile are required") {
			return fmt.Errorf("the ollama server must be updated to use `ollama create` with this client")
		}
211
212
213
214
215
216
		return err
	}

	return nil
}

217
218
func createBlob(cmd *cobra.Command, client *api.Client, path string, digest string, p *progress.Progress) (string, error) {
	realPath, err := filepath.EvalSymlinks(path)
Michael Yang's avatar
Michael Yang committed
219
220
221
222
	if err != nil {
		return "", err
	}

223
	bin, err := os.Open(realPath)
224
225
226
227
228
	if err != nil {
		return "", err
	}
	defer bin.Close()

229
230
231
232
233
234
235
236
	// Get file info to retrieve the size
	fileInfo, err := bin.Stat()
	if err != nil {
		return "", err
	}
	fileSize := fileInfo.Size()

	var pw progressWriter
237
238
239
240
	status := fmt.Sprintf("copying file %s 0%%", digest)
	spinner := progress.NewSpinner(status)
	p.Add(status, spinner)
	defer spinner.Stop()
241
242
243
244
245
246
247
248
249
250

	done := make(chan struct{})
	defer close(done)

	go func() {
		ticker := time.NewTicker(60 * time.Millisecond)
		defer ticker.Stop()
		for {
			select {
			case <-ticker.C:
251
				spinner.SetMessage(fmt.Sprintf("copying file %s %d%%", digest, int(100*pw.n.Load()/fileSize)))
252
			case <-done:
253
				spinner.SetMessage(fmt.Sprintf("copying file %s 100%%", digest))
254
255
256
257
258
				return
			}
		}
	}()

259
	if err := client.CreateBlob(cmd.Context(), digest, io.TeeReader(bin, &pw)); err != nil {
260
261
262
263
264
		return "", err
	}
	return digest, nil
}

265
266
267
268
269
270
271
272
273
type progressWriter struct {
	n atomic.Int64
}

func (w *progressWriter) Write(p []byte) (n int, err error) {
	w.n.Add(int64(len(p)))
	return len(p), nil
}

Patrick Devine's avatar
Patrick Devine committed
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
func loadOrUnloadModel(cmd *cobra.Command, opts *runOptions) error {
	p := progress.NewProgress(os.Stderr)
	defer p.StopAndClear()

	spinner := progress.NewSpinner("")
	p.Add("", spinner)

	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

	req := &api.GenerateRequest{
		Model:     opts.Model,
		KeepAlive: opts.KeepAlive,
289
290
291

		// pass Think here so we fail before getting to the chat prompt if the model doesn't support it
		Think: opts.Think,
Patrick Devine's avatar
Patrick Devine committed
292
293
	}

294
295
296
297
298
299
300
301
302
303
304
	return client.Generate(cmd.Context(), req, func(r api.GenerateResponse) error {
		if r.RemoteModel != "" && opts.ShowConnect {
			p.StopAndClear()
			if strings.HasPrefix(r.RemoteHost, "https://ollama.com") {
				fmt.Fprintf(os.Stderr, "Connecting to '%s' on 'ollama.com' ⚡\n", r.RemoteModel)
			} else {
				fmt.Fprintf(os.Stderr, "Connecting to '%s' on '%s'\n", r.RemoteModel, r.RemoteHost)
			}
		}
		return nil
	})
Patrick Devine's avatar
Patrick Devine committed
305
306
307
308
309
310
311
312
313
314
315
}

func StopHandler(cmd *cobra.Command, args []string) error {
	opts := &runOptions{
		Model:     args[0],
		KeepAlive: &api.Duration{Duration: 0},
	}
	if err := loadOrUnloadModel(cmd, opts); err != nil {
		if strings.Contains(err.Error(), "not found") {
			return fmt.Errorf("couldn't find model \"%s\" to stop", args[0])
		}
316
		return err
Patrick Devine's avatar
Patrick Devine committed
317
318
319
320
	}
	return nil
}

321
func RunHandler(cmd *cobra.Command, args []string) error {
322
323
324
	interactive := true

	opts := runOptions{
325
326
327
328
		Model:       args[0],
		WordWrap:    os.Getenv("TERM") == "xterm-256color",
		Options:     map[string]any{},
		ShowConnect: true,
329
330
331
332
333
334
335
336
	}

	format, err := cmd.Flags().GetString("format")
	if err != nil {
		return err
	}
	opts.Format = format

337
338
	thinkFlag := cmd.Flags().Lookup("think")
	if thinkFlag.Changed {
Michael Yang's avatar
Michael Yang committed
339
		thinkStr, err := cmd.Flags().GetString("think")
340
341
342
		if err != nil {
			return err
		}
Michael Yang's avatar
Michael Yang committed
343
344
345
346
347
348
349
350
351
352
353
354
355

		// Handle different values for --think
		switch thinkStr {
		case "", "true":
			// --think or --think=true
			opts.Think = &api.ThinkValue{Value: true}
		case "false":
			opts.Think = &api.ThinkValue{Value: false}
		case "high", "medium", "low":
			opts.Think = &api.ThinkValue{Value: thinkStr}
		default:
			return fmt.Errorf("invalid value for --think: %q (must be true, false, high, medium, or low)", thinkStr)
		}
356
357
358
359
360
361
362
363
364
	} else {
		opts.Think = nil
	}
	hidethinking, err := cmd.Flags().GetBool("hidethinking")
	if err != nil {
		return err
	}
	opts.HideThinking = hidethinking

365
366
367
368
369
370
371
372
373
374
375
376
	keepAlive, err := cmd.Flags().GetString("keepalive")
	if err != nil {
		return err
	}
	if keepAlive != "" {
		d, err := time.ParseDuration(keepAlive)
		if err != nil {
			return err
		}
		opts.KeepAlive = &api.Duration{Duration: d}
	}

377
378
379
380
381
382
383
384
385
	prompts := args[1:]
	// prepend stdin to the prompt if provided
	if !term.IsTerminal(int(os.Stdin.Fd())) {
		in, err := io.ReadAll(os.Stdin)
		if err != nil {
			return err
		}

		prompts = append([]string{string(in)}, prompts...)
386
		opts.ShowConnect = false
387
388
389
390
391
392
393
		opts.WordWrap = false
		interactive = false
	}
	opts.Prompt = strings.Join(prompts, " ")
	if len(prompts) > 0 {
		interactive = false
	}
394
395
396
397
	// Be quiet if we're redirecting to a pipe or file
	if !term.IsTerminal(int(os.Stdout.Fd())) {
		interactive = false
	}
398
399
400
401
402
403
404

	nowrap, err := cmd.Flags().GetBool("nowordwrap")
	if err != nil {
		return err
	}
	opts.WordWrap = !nowrap

405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
	// Fill out the rest of the options based on information about the
	// model.
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

	name := args[0]
	info, err := func() (*api.ShowResponse, error) {
		showReq := &api.ShowRequest{Name: name}
		info, err := client.Show(cmd.Context(), showReq)
		var se api.StatusError
		if errors.As(err, &se) && se.StatusCode == http.StatusNotFound {
			if err := PullHandler(cmd, []string{name}); err != nil {
				return nil, err
			}
			return client.Show(cmd.Context(), &api.ShowRequest{Name: name})
		}
		return info, err
	}()
	if err != nil {
		return err
427
428
	}

429
430
431
432
433
	opts.Think, err = inferThinkingOption(&info.Capabilities, &opts, thinkFlag.Changed)
	if err != nil {
		return err
	}

434
435
436
437
438
	opts.MultiModal = slices.Contains(info.Capabilities, model.CapabilityVision)

	// TODO: remove the projector info and vision info checks below,
	// these are left in for backwards compatibility with older servers
	// that don't have the capabilities field in the model info
439
440
441
442
443
444
445
446
447
448
	if len(info.ProjectorInfo) != 0 {
		opts.MultiModal = true
	}
	for k := range info.ModelInfo {
		if strings.Contains(k, ".vision.") {
			opts.MultiModal = true
			break
		}
	}

449
450
451
	opts.ParentModel = info.Details.ParentModel

	if interactive {
Patrick Devine's avatar
Patrick Devine committed
452
		if err := loadOrUnloadModel(cmd, &opts); err != nil {
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
			var sErr api.AuthorizationError
			if errors.As(err, &sErr) && sErr.StatusCode == http.StatusUnauthorized {
				pubKey, pkErr := auth.GetPublicKey()
				if pkErr != nil {
					return pkErr
				}
				// the server and the client both have the same public key
				if pubKey == sErr.PublicKey {
					h, _ := os.Hostname()
					encKey := base64.RawURLEncoding.EncodeToString([]byte(pubKey))
					fmt.Printf("You need to be signed in to Ollama to run Cloud models.\n\n")
					fmt.Printf(ConnectInstructions, url.PathEscape(h), encKey)
				}
				return nil
			}
Michael Yang's avatar
Michael Yang committed
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
			return err
		}

		for _, msg := range info.Messages {
			switch msg.Role {
			case "user":
				fmt.Printf(">>> %s\n", msg.Content)
			case "assistant":
				state := &displayResponseState{}
				displayResponse(msg.Content, opts.WordWrap, state)
				fmt.Println()
				fmt.Println()
			}
		}

483
484
485
		return generateInteractive(cmd, opts)
	}
	return generate(cmd, opts)
Bruce MacDonald's avatar
Bruce MacDonald committed
486
487
}

488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
func SigninHandler(cmd *cobra.Command, args []string) error {
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

	user, err := client.Whoami(cmd.Context())
	if err != nil {
		return err
	}

	if user != nil && user.Name != "" {
		fmt.Printf("You are already signed in as user '%s'\n", user.Name)
		fmt.Println()
		return nil
	}

	pubKey, pkErr := auth.GetPublicKey()
	if pkErr != nil {
		return pkErr
	}
	encKey := base64.RawURLEncoding.EncodeToString([]byte(pubKey))

	h, _ := os.Hostname()
	fmt.Printf(ConnectInstructions, url.PathEscape(h), encKey)

	return nil
}

func SignoutHandler(cmd *cobra.Command, args []string) error {
	pubKey, pkErr := auth.GetPublicKey()
	if pkErr != nil {
		return pkErr
	}
	encKey := base64.RawURLEncoding.EncodeToString([]byte(pubKey))

	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

	err = client.Signout(cmd.Context(), encKey)
	if err != nil {
		return err
	}
	fmt.Println("You have signed out of ollama.com")
	fmt.Println()
	return nil
}

538
func PushHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
539
	client, err := api.ClientFromEnvironment()
540
541
542
	if err != nil {
		return err
	}
543

544
545
546
547
548
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
549
550
551
552
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)
553
554
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
555

556
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
557
		if resp.Digest != "" {
558
559
560
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
561
562
563

			bar, ok := bars[resp.Digest]
			if !ok {
564
				bar = progress.NewBar(fmt.Sprintf("pushing %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
565
566
567
568
569
570
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
571
572
573
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
574
575
576
577
578
579

			status = resp.Status
			spinner = progress.NewSpinner(status)
			p.Add(status, spinner)
		}

580
581
582
		return nil
	}

Michael Yang's avatar
Michael Yang committed
583
	request := api.PushRequest{Name: args[0], Insecure: insecure}
584
585

	n := model.ParseName(args[0])
Michael Yang's avatar
Michael Yang committed
586
	if err := client.Push(cmd.Context(), &request, fn); err != nil {
587
588
589
		if spinner != nil {
			spinner.Stop()
		}
590
591
		errStr := strings.ToLower(err.Error())
		if strings.Contains(errStr, "access denied") || strings.Contains(errStr, "unauthorized") {
592
593
			return errors.New("you are not authorized to push to this namespace, create the model under a namespace you own")
		}
Michael Yang's avatar
Michael Yang committed
594
595
596
		return err
	}

597
	p.Stop()
598
	spinner.Stop()
599
600
601
602
603
604
605
606

	destination := n.String()
	if strings.HasSuffix(n.Host, ".ollama.ai") || strings.HasSuffix(n.Host, ".ollama.com") {
		destination = "https://ollama.com/" + strings.TrimSuffix(n.DisplayShortest(), ":latest")
	}
	fmt.Printf("\nYou can find your model at:\n\n")
	fmt.Printf("\t%s\n", destination)

Michael Yang's avatar
Michael Yang committed
607
	return nil
608
609
}

610
func ListHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
611
	client, err := api.ClientFromEnvironment()
612
613
614
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
615

Michael Yang's avatar
Michael Yang committed
616
	models, err := client.List(cmd.Context())
Patrick Devine's avatar
Patrick Devine committed
617
618
619
620
621
622
623
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
624
		if len(args) == 0 || strings.HasPrefix(strings.ToLower(m.Name), strings.ToLower(args[0])) {
625
626
627
628
629
630
631
632
			var size string
			if m.RemoteModel != "" {
				size = "-"
			} else {
				size = format.HumanBytes(m.Size)
			}

			data = append(data, []string{m.Name, m.Digest[:12], size, format.HumanTime(m.ModifiedAt, "Never")})
Michael Yang's avatar
Michael Yang committed
633
		}
Patrick Devine's avatar
Patrick Devine committed
634
635
636
	}

	table := tablewriter.NewWriter(os.Stdout)
Patrick Devine's avatar
Patrick Devine committed
637
	table.SetHeader([]string{"NAME", "ID", "SIZE", "MODIFIED"})
Patrick Devine's avatar
Patrick Devine committed
638
639
640
641
642
	table.SetHeaderAlignment(tablewriter.ALIGN_LEFT)
	table.SetAlignment(tablewriter.ALIGN_LEFT)
	table.SetHeaderLine(false)
	table.SetBorder(false)
	table.SetNoWhiteSpace(true)
Michael Yang's avatar
Michael Yang committed
643
	table.SetTablePadding("    ")
Patrick Devine's avatar
Patrick Devine committed
644
645
646
647
648
649
	table.AppendBulk(data)
	table.Render()

	return nil
}

650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
func ListRunningHandler(cmd *cobra.Command, args []string) error {
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

	models, err := client.ListRunning(cmd.Context())
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
		if len(args) == 0 || strings.HasPrefix(m.Name, args[0]) {
			var procStr string
			switch {
			case m.SizeVRAM == 0:
				procStr = "100% CPU"
			case m.SizeVRAM == m.Size:
				procStr = "100% GPU"
			case m.SizeVRAM > m.Size || m.Size == 0:
				procStr = "Unknown"
			default:
				sizeCPU := m.Size - m.SizeVRAM
				cpuPercent := math.Round(float64(sizeCPU) / float64(m.Size) * 100)
				procStr = fmt.Sprintf("%d%%/%d%% CPU/GPU", int(cpuPercent), int(100-cpuPercent))
			}
Patrick Devine's avatar
Patrick Devine committed
678
679
680
681
682
683
684
685

			var until string
			delta := time.Since(m.ExpiresAt)
			if delta > 0 {
				until = "Stopping..."
			} else {
				until = format.HumanTime(m.ExpiresAt, "Never")
			}
686
687
			ctxStr := strconv.Itoa(m.ContextLength)
			data = append(data, []string{m.Name, m.Digest[:12], format.HumanBytes(m.Size), procStr, ctxStr, until})
688
689
690
691
		}
	}

	table := tablewriter.NewWriter(os.Stdout)
692
	table.SetHeader([]string{"NAME", "ID", "SIZE", "PROCESSOR", "CONTEXT", "UNTIL"})
693
694
695
696
697
	table.SetHeaderAlignment(tablewriter.ALIGN_LEFT)
	table.SetAlignment(tablewriter.ALIGN_LEFT)
	table.SetHeaderLine(false)
	table.SetBorder(false)
	table.SetNoWhiteSpace(true)
Michael Yang's avatar
Michael Yang committed
698
	table.SetTablePadding("    ")
699
700
701
702
703
704
	table.AppendBulk(data)
	table.Render()

	return nil
}

705
func DeleteHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
706
	client, err := api.ClientFromEnvironment()
707
708
709
	if err != nil {
		return err
	}
710

711
712
713
714
715
716
	// Unload the model if it's running before deletion
	opts := &runOptions{
		Model:     args[0],
		KeepAlive: &api.Duration{Duration: 0},
	}
	if err := loadOrUnloadModel(cmd, opts); err != nil {
717
718
		if !strings.Contains(strings.ToLower(err.Error()), "not found") {
			fmt.Fprintf(os.Stderr, "Warning: unable to stop model '%s'\n", args[0])
719
720
721
		}
	}

722
723
	for _, name := range args {
		req := api.DeleteRequest{Name: name}
Michael Yang's avatar
Michael Yang committed
724
		if err := client.Delete(cmd.Context(), &req); err != nil {
725
726
727
			return err
		}
		fmt.Printf("deleted '%s'\n", name)
728
729
730
731
	}
	return nil
}

Patrick Devine's avatar
Patrick Devine committed
732
func ShowHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
733
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
734
735
736
737
738
739
740
741
742
	if err != nil {
		return err
	}

	license, errLicense := cmd.Flags().GetBool("license")
	modelfile, errModelfile := cmd.Flags().GetBool("modelfile")
	parameters, errParams := cmd.Flags().GetBool("parameters")
	system, errSystem := cmd.Flags().GetBool("system")
	template, errTemplate := cmd.Flags().GetBool("template")
743
	verbose, errVerbose := cmd.Flags().GetBool("verbose")
Patrick Devine's avatar
Patrick Devine committed
744

745
	for _, boolErr := range []error{errLicense, errModelfile, errParams, errSystem, errTemplate, errVerbose} {
Patrick Devine's avatar
Patrick Devine committed
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
		if boolErr != nil {
			return errors.New("error retrieving flags")
		}
	}

	flagsSet := 0
	showType := ""

	if license {
		flagsSet++
		showType = "license"
	}

	if modelfile {
		flagsSet++
		showType = "modelfile"
	}

	if parameters {
		flagsSet++
		showType = "parameters"
	}

	if system {
		flagsSet++
		showType = "system"
	}

	if template {
		flagsSet++
		showType = "template"
	}

	if flagsSet > 1 {
780
		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
781
782
	}

783
	req := api.ShowRequest{Name: args[0], Verbose: verbose}
784
785
786
787
	resp, err := client.Show(cmd.Context(), &req)
	if err != nil {
		return err
	}
788

789
	if flagsSet == 1 {
790
791
792
793
794
795
796
797
		switch showType {
		case "license":
			fmt.Println(resp.License)
		case "modelfile":
			fmt.Println(resp.Modelfile)
		case "parameters":
			fmt.Println(resp.Parameters)
		case "system":
798
			fmt.Print(resp.System)
799
		case "template":
800
			fmt.Print(resp.Template)
801
802
803
		}

		return nil
Patrick Devine's avatar
Patrick Devine committed
804
805
	}

806
	return showInfo(resp, verbose, os.Stdout)
807
808
}

809
func showInfo(resp *api.ShowResponse, verbose bool, w io.Writer) error {
Michael Yang's avatar
Michael Yang committed
810
811
812
813
814
815
816
	tableRender := func(header string, rows func() [][]string) {
		fmt.Fprintln(w, " ", header)
		table := tablewriter.NewWriter(w)
		table.SetAlignment(tablewriter.ALIGN_LEFT)
		table.SetBorder(false)
		table.SetNoWhiteSpace(true)
		table.SetTablePadding("    ")
817

Michael Yang's avatar
Michael Yang committed
818
819
820
		switch header {
		case "Template", "System", "License":
			table.SetColWidth(100)
821
822
		}

Michael Yang's avatar
Michael Yang committed
823
824
825
		table.AppendBulk(rows())
		table.Render()
		fmt.Fprintln(w)
Patrick Devine's avatar
Patrick Devine committed
826
827
	}

Michael Yang's avatar
Michael Yang committed
828
	tableRender("Model", func() (rows [][]string) {
829
830
831
832
833
		if resp.RemoteHost != "" {
			rows = append(rows, []string{"", "Remote model", resp.RemoteModel})
			rows = append(rows, []string{"", "Remote URL", resp.RemoteHost})
		}

Michael Yang's avatar
Michael Yang committed
834
835
836
		if resp.ModelInfo != nil {
			arch := resp.ModelInfo["general.architecture"].(string)
			rows = append(rows, []string{"", "architecture", arch})
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858

			var paramStr string
			if resp.Details.ParameterSize != "" {
				paramStr = resp.Details.ParameterSize
			} else if v, ok := resp.ModelInfo["general.parameter_count"]; ok {
				if f, ok := v.(float64); ok {
					paramStr = format.HumanNumber(uint64(f))
				}
			}
			rows = append(rows, []string{"", "parameters", paramStr})

			if v, ok := resp.ModelInfo[fmt.Sprintf("%s.context_length", arch)]; ok {
				if f, ok := v.(float64); ok {
					rows = append(rows, []string{"", "context length", strconv.FormatFloat(f, 'f', -1, 64)})
				}
			}

			if v, ok := resp.ModelInfo[fmt.Sprintf("%s.embedding_length", arch)]; ok {
				if f, ok := v.(float64); ok {
					rows = append(rows, []string{"", "embedding length", strconv.FormatFloat(f, 'f', -1, 64)})
				}
			}
Michael Yang's avatar
Michael Yang committed
859
860
861
862
863
864
865
		} else {
			rows = append(rows, []string{"", "architecture", resp.Details.Family})
			rows = append(rows, []string{"", "parameters", resp.Details.ParameterSize})
		}
		rows = append(rows, []string{"", "quantization", resp.Details.QuantizationLevel})
		return
	})
866

867
868
869
870
871
872
873
874
875
	if len(resp.Capabilities) > 0 {
		tableRender("Capabilities", func() (rows [][]string) {
			for _, capability := range resp.Capabilities {
				rows = append(rows, []string{"", capability.String()})
			}
			return
		})
	}

Michael Yang's avatar
Michael Yang committed
876
877
878
879
880
881
882
883
884
	if resp.ProjectorInfo != nil {
		tableRender("Projector", func() (rows [][]string) {
			arch := resp.ProjectorInfo["general.architecture"].(string)
			rows = append(rows, []string{"", "architecture", arch})
			rows = append(rows, []string{"", "parameters", format.HumanNumber(uint64(resp.ProjectorInfo["general.parameter_count"].(float64)))})
			rows = append(rows, []string{"", "embedding length", strconv.FormatFloat(resp.ProjectorInfo[fmt.Sprintf("%s.vision.embedding_length", arch)].(float64), 'f', -1, 64)})
			rows = append(rows, []string{"", "dimensions", strconv.FormatFloat(resp.ProjectorInfo[fmt.Sprintf("%s.vision.projection_dim", arch)].(float64), 'f', -1, 64)})
			return
		})
885
886
	}

Michael Yang's avatar
Michael Yang committed
887
888
889
890
891
892
893
894
895
896
	if resp.Parameters != "" {
		tableRender("Parameters", func() (rows [][]string) {
			scanner := bufio.NewScanner(strings.NewReader(resp.Parameters))
			for scanner.Scan() {
				if text := scanner.Text(); text != "" {
					rows = append(rows, append([]string{""}, strings.Fields(text)...))
				}
			}
			return
		})
897
898
	}

899
900
901
902
903
904
905
906
907
908
909
	if resp.ModelInfo != nil && verbose {
		tableRender("Metadata", func() (rows [][]string) {
			keys := make([]string, 0, len(resp.ModelInfo))
			for k := range resp.ModelInfo {
				keys = append(keys, k)
			}
			sort.Strings(keys)

			for _, k := range keys {
				var v string
				switch vData := resp.ModelInfo[k].(type) {
910
911
				case bool:
					v = fmt.Sprintf("%t", vData)
912
913
914
915
916
				case string:
					v = vData
				case float64:
					v = fmt.Sprintf("%g", vData)
				case []any:
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
					targetWidth := 10 // Small width where we are displaying the data in a column

					var itemsToShow int
					totalWidth := 1 // Start with 1 for opening bracket

					// Find how many we can fit
					for i := range vData {
						itemStr := fmt.Sprintf("%v", vData[i])
						width := runewidth.StringWidth(itemStr)

						// Add separator width (", ") for all items except the first
						if i > 0 {
							width += 2
						}

						// Check if adding this item would exceed our width limit
						if totalWidth+width > targetWidth && i > 0 {
							break
						}

						totalWidth += width
						itemsToShow++
					}

					// Format the output
					if itemsToShow < len(vData) {
						v = fmt.Sprintf("%v", vData[:itemsToShow])
						v = strings.TrimSuffix(v, "]")
						v += fmt.Sprintf(" ...+%d more]", len(vData)-itemsToShow)
					} else {
						v = fmt.Sprintf("%v", vData)
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
					}
				default:
					v = fmt.Sprintf("%T", vData)
				}
				rows = append(rows, []string{"", k, v})
			}
			return
		})
	}

	if len(resp.Tensors) > 0 && verbose {
		tableRender("Tensors", func() (rows [][]string) {
			for _, t := range resp.Tensors {
				rows = append(rows, []string{"", t.Name, t.Type, fmt.Sprint(t.Shape)})
			}
			return
		})
	}

Michael Yang's avatar
Michael Yang committed
967
968
	head := func(s string, n int) (rows [][]string) {
		scanner := bufio.NewScanner(strings.NewReader(s))
969
970
971
972
973
		count := 0
		for scanner.Scan() {
			text := strings.TrimSpace(scanner.Text())
			if text == "" {
				continue
974
			}
975
976
977
978
979
980
981
			count++
			if n < 0 || count <= n {
				rows = append(rows, []string{"", text})
			}
		}
		if n >= 0 && count > n {
			rows = append(rows, []string{"", "..."})
982
		}
Michael Yang's avatar
Michael Yang committed
983
		return
984
985
	}

Michael Yang's avatar
Michael Yang committed
986
987
988
989
990
	if resp.System != "" {
		tableRender("System", func() [][]string {
			return head(resp.System, 2)
		})
	}
991

Michael Yang's avatar
Michael Yang committed
992
993
994
995
	if resp.License != "" {
		tableRender("License", func() [][]string {
			return head(resp.License, 2)
		})
996
	}
Michael Yang's avatar
Michael Yang committed
997
998

	return nil
999
1000
}

Patrick Devine's avatar
Patrick Devine committed
1001
func CopyHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
1002
	client, err := api.ClientFromEnvironment()
1003
1004
1005
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
1006
1007

	req := api.CopyRequest{Source: args[0], Destination: args[1]}
Michael Yang's avatar
Michael Yang committed
1008
	if err := client.Copy(cmd.Context(), &req); err != nil {
Patrick Devine's avatar
Patrick Devine committed
1009
1010
1011
1012
1013
1014
		return err
	}
	fmt.Printf("copied '%s' to '%s'\n", args[0], args[1])
	return nil
}

1015
func PullHandler(cmd *cobra.Command, args []string) error {
1016
1017
1018
1019
1020
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
1021
	client, err := api.ClientFromEnvironment()
1022
1023
1024
	if err != nil {
		return err
	}
1025

Michael Yang's avatar
Michael Yang committed
1026
1027
1028
1029
1030
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)

1031
1032
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
1033

1034
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
1035
		if resp.Digest != "" {
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
			if resp.Completed == 0 {
				// This is the initial status update for the
				// layer, which the server sends before
				// beginning the download, for clients to
				// compute total size and prepare for
				// downloads, if needed.
				//
				// Skipping this here to avoid showing a 0%
				// progress bar, which *should* clue the user
				// into the fact that many things are being
				// downloaded and that the current active
				// download is not that last. However, in rare
				// cases it seems to be triggering to some, and
				// it isn't worth explaining, so just ignore
				// and regress to the old UI that keeps giving
				// you the "But wait, there is more!" after
				// each "100% done" bar, which is "better."
				return nil
			}

1056
1057
1058
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
1059
1060
1061

			bar, ok := bars[resp.Digest]
			if !ok {
1062
1063
1064
1065
1066
1067
				name, isDigest := strings.CutPrefix(resp.Digest, "sha256:")
				name = strings.TrimSpace(name)
				if isDigest {
					name = name[:min(12, len(name))]
				}
				bar = progress.NewBar(fmt.Sprintf("pulling %s:", name), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
1068
1069
1070
1071
1072
1073
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
1074
1075
1076
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
1077
1078
1079
1080
1081
1082

			status = resp.Status
			spinner = progress.NewSpinner(status)
			p.Add(status, spinner)
		}

1083
1084
		return nil
	}
1085

Michael Yang's avatar
Michael Yang committed
1086
	request := api.PullRequest{Name: args[0], Insecure: insecure}
1087
	return client.Pull(cmd.Context(), &request, fn)
Michael Yang's avatar
Michael Yang committed
1088
1089
}

1090
1091
type generateContextKey string

1092
type runOptions struct {
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
	Model        string
	ParentModel  string
	Prompt       string
	Messages     []api.Message
	WordWrap     bool
	Format       string
	System       string
	Images       []api.ImageData
	Options      map[string]any
	MultiModal   bool
	KeepAlive    *api.Duration
Michael Yang's avatar
Michael Yang committed
1104
	Think        *api.ThinkValue
1105
	HideThinking bool
1106
	ShowConnect  bool
1107
1108
}

1109
1110
1111
1112
1113
1114
1115
1116
1117
type displayResponseState struct {
	lineLength int
	wordBuffer string
}

func displayResponse(content string, wordWrap bool, state *displayResponseState) {
	termWidth, _, _ := term.GetSize(int(os.Stdout.Fd()))
	if wordWrap && termWidth >= 10 {
		for _, ch := range content {
Josh Yan's avatar
Josh Yan committed
1118
1119
			if state.lineLength+1 > termWidth-5 {
				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
1120
1121
1122
1123
1124
1125
1126
					fmt.Printf("%s%c", state.wordBuffer, ch)
					state.wordBuffer = ""
					state.lineLength = 0
					continue
				}

				// backtrack the length of the last word and clear to the end of the line
1127
1128
				a := runewidth.StringWidth(state.wordBuffer)
				if a > 0 {
1129
					fmt.Printf("\x1b[%dD", a)
1130
1131
				}
				fmt.Printf("\x1b[K\n")
1132
				fmt.Printf("%s%c", state.wordBuffer, ch)
1133
1134
1135
				chWidth := runewidth.RuneWidth(ch)

				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
1136
1137
			} else {
				fmt.Print(string(ch))
1138
1139
1140
1141
				state.lineLength += runewidth.RuneWidth(ch)
				if runewidth.RuneWidth(ch) >= 2 {
					state.wordBuffer = ""
					continue
Josh Yan's avatar
Josh Yan committed
1142
				}
1143
1144

				switch ch {
Michael Yang's avatar
Michael Yang committed
1145
				case ' ', '\t':
1146
					state.wordBuffer = ""
Michael Yang's avatar
Michael Yang committed
1147
				case '\n', '\r':
1148
					state.lineLength = 0
Michael Yang's avatar
Michael Yang committed
1149
					state.wordBuffer = ""
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
				default:
					state.wordBuffer += string(ch)
				}
			}
		}
	} else {
		fmt.Printf("%s%s", state.wordBuffer, content)
		if len(state.wordBuffer) > 0 {
			state.wordBuffer = ""
		}
	}
}

1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
func thinkingOutputOpeningText(plainText bool) string {
	text := "Thinking...\n"

	if plainText {
		return text
	}

	return readline.ColorGrey + readline.ColorBold + text + readline.ColorDefault + readline.ColorGrey
}

func thinkingOutputClosingText(plainText bool) string {
	text := "...done thinking.\n\n"

	if plainText {
		return text
	}

	return readline.ColorGrey + readline.ColorBold + text + readline.ColorDefault
}

1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
func chat(cmd *cobra.Command, opts runOptions) (*api.Message, error) {
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return nil, err
	}

	p := progress.NewProgress(os.Stderr)
	defer p.StopAndClear()

	spinner := progress.NewSpinner("")
	p.Add("", spinner)

	cancelCtx, cancel := context.WithCancel(cmd.Context())
	defer cancel()

	sigChan := make(chan os.Signal, 1)
	signal.Notify(sigChan, syscall.SIGINT)

	go func() {
		<-sigChan
		cancel()
	}()

	var state *displayResponseState = &displayResponseState{}
Michael Yang's avatar
Michael Yang committed
1207
	var thinkingContent strings.Builder
1208
1209
	var latest api.ChatResponse
	var fullResponse strings.Builder
1210
1211
	var thinkTagOpened bool = false
	var thinkTagClosed bool = false
1212

1213
1214
	role := "assistant"

1215
	fn := func(response api.ChatResponse) error {
1216
1217
1218
		if response.Message.Content != "" || !opts.HideThinking {
			p.StopAndClear()
		}
1219
1220
1221
1222

		latest = response

		role = response.Message.Role
1223
1224
1225
1226
		if response.Message.Thinking != "" && !opts.HideThinking {
			if !thinkTagOpened {
				fmt.Print(thinkingOutputOpeningText(false))
				thinkTagOpened = true
Michael Yang's avatar
Michael Yang committed
1227
				thinkTagClosed = false
1228
			}
Michael Yang's avatar
Michael Yang committed
1229
			thinkingContent.WriteString(response.Message.Thinking)
1230
1231
1232
			displayResponse(response.Message.Thinking, opts.WordWrap, state)
		}

1233
		content := response.Message.Content
Michael Yang's avatar
Michael Yang committed
1234
1235
1236
1237
		if thinkTagOpened && !thinkTagClosed && (content != "" || len(response.Message.ToolCalls) > 0) {
			if !strings.HasSuffix(thinkingContent.String(), "\n") {
				fmt.Println()
			}
1238
			fmt.Print(thinkingOutputClosingText(false))
Michael Yang's avatar
Michael Yang committed
1239
			thinkTagOpened = false
1240
			thinkTagClosed = true
Michael Yang's avatar
Michael Yang committed
1241
			state = &displayResponseState{}
1242
1243
1244
1245
1246
		}
		// purposefully not putting thinking blocks in the response, which would
		// only be needed if we later added tool calling to the cli (they get
		// filtered out anyway since current models don't expect them unless you're
		// about to finish some tool calls)
1247
1248
		fullResponse.WriteString(content)

Michael Yang's avatar
Michael Yang committed
1249
1250
1251
1252
1253
1254
1255
		if response.Message.ToolCalls != nil {
			toolCalls := response.Message.ToolCalls
			if len(toolCalls) > 0 {
				fmt.Print(renderToolCalls(toolCalls, false))
			}
		}

1256
1257
1258
1259
1260
		displayResponse(content, opts.WordWrap, state)

		return nil
	}

1261
1262
1263
1264
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

1265
1266
1267
	req := &api.ChatRequest{
		Model:    opts.Model,
		Messages: opts.Messages,
1268
		Format:   json.RawMessage(opts.Format),
1269
		Options:  opts.Options,
1270
		Think:    opts.Think,
1271
1272
	}

1273
1274
1275
1276
	if opts.KeepAlive != nil {
		req.KeepAlive = opts.KeepAlive
	}

1277
1278
1279
1280
	if err := client.Chat(cancelCtx, req, fn); err != nil {
		if errors.Is(err, context.Canceled) {
			return nil, nil
		}
1281
1282
1283
1284
1285
1286
1287
1288

		// this error should ideally be wrapped properly by the client
		if strings.Contains(err.Error(), "upstream error") {
			p.StopAndClear()
			fmt.Println("An error occurred while processing your message. Please try again.")
			fmt.Println()
			return nil, nil
		}
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
		return nil, err
	}

	if len(opts.Messages) > 0 {
		fmt.Println()
		fmt.Println()
	}

	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
		return nil, err
	}

	if verbose {
		latest.Summary()
	}

	return &api.Message{Role: role, Content: fullResponse.String()}, nil
}

func generate(cmd *cobra.Command, opts runOptions) error {
Michael Yang's avatar
Michael Yang committed
1310
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
1311
	if err != nil {
1312
		return err
Patrick Devine's avatar
Patrick Devine committed
1313
	}
Michael Yang's avatar
Michael Yang committed
1314

Michael Yang's avatar
Michael Yang committed
1315
	p := progress.NewProgress(os.Stderr)
1316
	defer p.StopAndClear()
1317

Michael Yang's avatar
Michael Yang committed
1318
1319
1320
	spinner := progress.NewSpinner("")
	p.Add("", spinner)

1321
1322
1323
1324
1325
1326
1327
	var latest api.GenerateResponse

	generateContext, ok := cmd.Context().Value(generateContextKey("context")).([]int)
	if !ok {
		generateContext = []int{}
	}

Michael Yang's avatar
Michael Yang committed
1328
	ctx, cancel := context.WithCancel(cmd.Context())
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
	defer cancel()

	sigChan := make(chan os.Signal, 1)
	signal.Notify(sigChan, syscall.SIGINT)

	go func() {
		<-sigChan
		cancel()
	}()

1339
	var state *displayResponseState = &displayResponseState{}
Michael Yang's avatar
Michael Yang committed
1340
	var thinkingContent strings.Builder
1341
1342
	var thinkTagOpened bool = false
	var thinkTagClosed bool = false
1343

1344
	plainText := !term.IsTerminal(int(os.Stdout.Fd()))
1345

1346
	fn := func(response api.GenerateResponse) error {
Patrick Devine's avatar
Patrick Devine committed
1347
		latest = response
1348
		content := response.Response
1349

1350
1351
1352
1353
1354
1355
1356
1357
		if response.Response != "" || !opts.HideThinking {
			p.StopAndClear()
		}

		if response.Thinking != "" && !opts.HideThinking {
			if !thinkTagOpened {
				fmt.Print(thinkingOutputOpeningText(plainText))
				thinkTagOpened = true
Michael Yang's avatar
Michael Yang committed
1358
				thinkTagClosed = false
1359
			}
Michael Yang's avatar
Michael Yang committed
1360
			thinkingContent.WriteString(response.Thinking)
1361
1362
1363
			displayResponse(response.Thinking, opts.WordWrap, state)
		}

Michael Yang's avatar
Michael Yang committed
1364
1365
1366
1367
		if thinkTagOpened && !thinkTagClosed && (content != "" || len(response.ToolCalls) > 0) {
			if !strings.HasSuffix(thinkingContent.String(), "\n") {
				fmt.Println()
			}
1368
			fmt.Print(thinkingOutputClosingText(plainText))
Michael Yang's avatar
Michael Yang committed
1369
			thinkTagOpened = false
1370
			thinkTagClosed = true
Michael Yang's avatar
Michael Yang committed
1371
			state = &displayResponseState{}
1372
1373
		}

1374
		displayResponse(content, opts.WordWrap, state)
1375

Michael Yang's avatar
Michael Yang committed
1376
1377
1378
1379
1380
1381
1382
		if response.ToolCalls != nil {
			toolCalls := response.ToolCalls
			if len(toolCalls) > 0 {
				fmt.Print(renderToolCalls(toolCalls, plainText))
			}
		}

Patrick Devine's avatar
Patrick Devine committed
1383
1384
		return nil
	}
1385

1386
1387
1388
1389
1390
1391
1392
	if opts.MultiModal {
		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
		if err != nil {
			return err
		}
	}

1393
1394
1395
1396
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

Michael Yang's avatar
Michael Yang committed
1397
	request := api.GenerateRequest{
1398
1399
1400
1401
		Model:     opts.Model,
		Prompt:    opts.Prompt,
		Context:   generateContext,
		Images:    opts.Images,
1402
		Format:    json.RawMessage(opts.Format),
1403
1404
1405
		System:    opts.System,
		Options:   opts.Options,
		KeepAlive: opts.KeepAlive,
1406
		Think:     opts.Think,
Michael Yang's avatar
Michael Yang committed
1407
1408
1409
	}

	if err := client.Generate(ctx, &request, fn); err != nil {
1410
		if errors.Is(err, context.Canceled) {
1411
			return nil
1412
		}
1413
		return err
Patrick Devine's avatar
Patrick Devine committed
1414
	}
1415

1416
	if opts.Prompt != "" {
Michael Yang's avatar
Michael Yang committed
1417
1418
		fmt.Println()
		fmt.Println()
Patrick Devine's avatar
Patrick Devine committed
1419
	}
1420

1421
1422
1423
1424
	if !latest.Done {
		return nil
	}

Patrick Devine's avatar
Patrick Devine committed
1425
1426
	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
1427
		return err
Patrick Devine's avatar
Patrick Devine committed
1428
	}
Michael Yang's avatar
Michael Yang committed
1429

Patrick Devine's avatar
Patrick Devine committed
1430
1431
	if verbose {
		latest.Summary()
Michael Yang's avatar
Michael Yang committed
1432
	}
Michael Yang's avatar
Michael Yang committed
1433

Patrick Devine's avatar
Patrick Devine committed
1434
1435
1436
	ctx = context.WithValue(cmd.Context(), generateContextKey("context"), latest.Context)
	cmd.SetContext(ctx)

1437
	return nil
Michael Yang's avatar
Michael Yang committed
1438
1439
}

1440
func RunServer(_ *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1441
	if err := initializeKeypair(); err != nil {
1442
1443
1444
		return err
	}

Michael Yang's avatar
host  
Michael Yang committed
1445
	ln, err := net.Listen("tcp", envconfig.Host().Host)
1446
1447
1448
	if err != nil {
		return err
	}
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1449

1450
1451
1452
1453
1454
1455
	err = server.Serve(ln)
	if errors.Is(err, http.ErrServerClosed) {
		return nil
	}

	return err
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1456
1457
}

1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
func initializeKeypair() error {
	home, err := os.UserHomeDir()
	if err != nil {
		return err
	}

	privKeyPath := filepath.Join(home, ".ollama", "id_ed25519")
	pubKeyPath := filepath.Join(home, ".ollama", "id_ed25519.pub")

	_, err = os.Stat(privKeyPath)
	if os.IsNotExist(err) {
		fmt.Printf("Couldn't find '%s'. Generating new private key.\n", privKeyPath)
Michael Yang's avatar
Michael Yang committed
1470
		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
1471
1472
1473
1474
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1475
		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
1476
1477
1478
1479
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1480
		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
1481
1482
1483
			return fmt.Errorf("could not create directory %w", err)
		}

Michael Yang's avatar
Michael Yang committed
1484
		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
1485
1486
1487
			return err
		}

Michael Yang's avatar
Michael Yang committed
1488
		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
1489
1490
1491
1492
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1493
		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
1494

Michael Yang's avatar
Michael Yang committed
1495
		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
1496
1497
1498
			return err
		}

Michael Yang's avatar
Michael Yang committed
1499
		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
1500
1501
1502
1503
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1504
func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1505
	client, err := api.ClientFromEnvironment()
1506
1507
1508
	if err != nil {
		return err
	}
Michael Yang's avatar
Michael Yang committed
1509
	if err := client.Heartbeat(cmd.Context()); err != nil {
1510
		if !(strings.Contains(err.Error(), " refused") || strings.Contains(err.Error(), "could not connect")) {
Bruce MacDonald's avatar
Bruce MacDonald committed
1511
1512
			return err
		}
1513
		if err := startApp(cmd.Context(), client); err != nil {
1514
			return fmt.Errorf("ollama server not responding - %w", err)
1515
1516
1517
1518
1519
		}
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1520
1521
1522
1523
1524
1525
1526
1527
func versionHandler(cmd *cobra.Command, _ []string) {
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return
	}

	serverVersion, err := client.Version(cmd.Context())
	if err != nil {
Michael Yang's avatar
Michael Yang committed
1528
1529
1530
1531
1532
		fmt.Println("Warning: could not connect to a running Ollama instance")
	}

	if serverVersion != "" {
		fmt.Printf("ollama version is %s\n", serverVersion)
Michael Yang's avatar
Michael Yang committed
1533
1534
	}

1535
	if serverVersion != version.Version {
Michael Yang's avatar
Michael Yang committed
1536
		fmt.Printf("Warning: client version is %s\n", version.Version)
1537
	}
Michael Yang's avatar
Michael Yang committed
1538
1539
}

1540
func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
1541
1542
1543
1544
1545
	if len(envs) == 0 {
		return
	}

	envUsage := `
1546
1547
Environment Variables:
`
1548
	for _, e := range envs {
1549
		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
1550
1551
1552
	}

	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
1553
1554
}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1555
1556
func NewCLI() *cobra.Command {
	log.SetFlags(log.LstdFlags | log.Lshortfile)
Michael Yang's avatar
Michael Yang committed
1557
	cobra.EnableCommandSorting = false
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1558

1559
	if runtime.GOOS == "windows" && term.IsTerminal(int(os.Stdout.Fd())) {
1560
		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
1561
1562
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1563
	rootCmd := &cobra.Command{
1564
1565
1566
1567
		Use:           "ollama",
		Short:         "Large language model runner",
		SilenceUsage:  true,
		SilenceErrors: true,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1568
1569
1570
		CompletionOptions: cobra.CompletionOptions{
			DisableDefaultCmd: true,
		},
Michael Yang's avatar
Michael Yang committed
1571
1572
1573
1574
1575
1576
1577
1578
		Run: func(cmd *cobra.Command, args []string) {
			if version, _ := cmd.Flags().GetBool("version"); version {
				versionHandler(cmd, args)
				return
			}

			cmd.Print(cmd.UsageString())
		},
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1579
1580
	}

Michael Yang's avatar
Michael Yang committed
1581
	rootCmd.Flags().BoolP("version", "v", false, "Show version information")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1582

1583
	createCmd := &cobra.Command{
1584
		Use:     "create MODEL",
1585
		Short:   "Create a model",
Michael Yang's avatar
Michael Yang committed
1586
		Args:    cobra.ExactArgs(1),
1587
1588
		PreRunE: checkServerHeartbeat,
		RunE:    CreateHandler,
1589
1590
	}

1591
	createCmd.Flags().StringP("file", "f", "", "Name of the Modelfile (default \"Modelfile\")")
1592
	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_K_M)")
1593

Patrick Devine's avatar
Patrick Devine committed
1594
1595
1596
	showCmd := &cobra.Command{
		Use:     "show MODEL",
		Short:   "Show information for a model",
Michael Yang's avatar
Michael Yang committed
1597
		Args:    cobra.ExactArgs(1),
Patrick Devine's avatar
Patrick Devine committed
1598
1599
1600
1601
1602
1603
1604
1605
		PreRunE: checkServerHeartbeat,
		RunE:    ShowHandler,
	}

	showCmd.Flags().Bool("license", false, "Show license of a model")
	showCmd.Flags().Bool("modelfile", false, "Show Modelfile of a model")
	showCmd.Flags().Bool("parameters", false, "Show parameters of a model")
	showCmd.Flags().Bool("template", false, "Show template of a model")
1606
	showCmd.Flags().Bool("system", false, "Show system message of a model")
1607
	showCmd.Flags().BoolP("verbose", "v", false, "Show detailed model information")
Patrick Devine's avatar
Patrick Devine committed
1608

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1609
	runCmd := &cobra.Command{
1610
1611
1612
1613
1614
		Use:     "run MODEL [PROMPT]",
		Short:   "Run a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    RunHandler,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1615
1616
	}

1617
	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
1618
	runCmd.Flags().Bool("verbose", false, "Show timings for response")
1619
	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
1620
	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1621
	runCmd.Flags().String("format", "", "Response format (e.g. json)")
Michael Yang's avatar
Michael Yang committed
1622
1623
	runCmd.Flags().String("think", "", "Enable thinking mode: true/false or high/medium/low for supported models")
	runCmd.Flags().Lookup("think").NoOptDefVal = "true"
1624
	runCmd.Flags().Bool("hidethinking", false, "Hide thinking output (if provided)")
Patrick Devine's avatar
Patrick Devine committed
1625
1626
1627
1628
1629
1630
1631
1632
1633

	stopCmd := &cobra.Command{
		Use:     "stop MODEL",
		Short:   "Stop a running model",
		Args:    cobra.ExactArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    StopHandler,
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1634
1635
1636
1637
	serveCmd := &cobra.Command{
		Use:     "serve",
		Aliases: []string{"start"},
		Short:   "Start ollama",
Michael Yang's avatar
Michael Yang committed
1638
		Args:    cobra.ExactArgs(0),
Michael Yang's avatar
Michael Yang committed
1639
		RunE:    RunServer,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1640
1641
	}

1642
	pullCmd := &cobra.Command{
1643
1644
		Use:     "pull MODEL",
		Short:   "Pull a model from a registry",
Michael Yang's avatar
Michael Yang committed
1645
		Args:    cobra.ExactArgs(1),
1646
1647
		PreRunE: checkServerHeartbeat,
		RunE:    PullHandler,
1648
1649
	}

1650
1651
	pullCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1652
	pushCmd := &cobra.Command{
1653
1654
		Use:     "push MODEL",
		Short:   "Push a model to a registry",
Michael Yang's avatar
Michael Yang committed
1655
		Args:    cobra.ExactArgs(1),
1656
1657
		PreRunE: checkServerHeartbeat,
		RunE:    PushHandler,
1658
1659
	}

1660
1661
	pushCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
	signinCmd := &cobra.Command{
		Use:     "signin",
		Short:   "Sign in to ollama.com",
		Args:    cobra.ExactArgs(0),
		PreRunE: checkServerHeartbeat,
		RunE:    SigninHandler,
	}

	signoutCmd := &cobra.Command{
		Use:     "signout",
		Short:   "Sign out from ollama.com",
		Args:    cobra.ExactArgs(0),
		PreRunE: checkServerHeartbeat,
		RunE:    SignoutHandler,
	}

Patrick Devine's avatar
Patrick Devine committed
1678
	listCmd := &cobra.Command{
1679
		Use:     "list",
Patrick Devine's avatar
Patrick Devine committed
1680
		Aliases: []string{"ls"},
1681
		Short:   "List models",
1682
		PreRunE: checkServerHeartbeat,
1683
		RunE:    ListHandler,
1684
	}
1685
1686
1687
1688
1689
1690
1691

	psCmd := &cobra.Command{
		Use:     "ps",
		Short:   "List running models",
		PreRunE: checkServerHeartbeat,
		RunE:    ListRunningHandler,
	}
Patrick Devine's avatar
Patrick Devine committed
1692
	copyCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1693
		Use:     "cp SOURCE DESTINATION",
1694
		Short:   "Copy a model",
Michael Yang's avatar
Michael Yang committed
1695
		Args:    cobra.ExactArgs(2),
1696
1697
		PreRunE: checkServerHeartbeat,
		RunE:    CopyHandler,
Patrick Devine's avatar
Patrick Devine committed
1698
1699
	}

1700
	deleteCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1701
		Use:     "rm MODEL [MODEL...]",
1702
1703
1704
1705
		Short:   "Remove a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    DeleteHandler,
Patrick Devine's avatar
Patrick Devine committed
1706
1707
	}

1708
1709
1710
1711
1712
1713
1714
1715
1716
1717
1718
1719
	runnerCmd := &cobra.Command{
		Use:    "runner",
		Hidden: true,
		RunE: func(cmd *cobra.Command, args []string) error {
			return runner.Execute(os.Args[1:])
		},
		FParseErrWhitelist: cobra.FParseErrWhitelist{UnknownFlags: true},
	}
	runnerCmd.SetHelpFunc(func(cmd *cobra.Command, args []string) {
		_ = runner.Execute(args[1:])
	})

1720
1721
1722
	envVars := envconfig.AsMap()

	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
1723

1724
1725
1726
1727
	for _, cmd := range []*cobra.Command{
		createCmd,
		showCmd,
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1728
		stopCmd,
1729
1730
1731
		pullCmd,
		pushCmd,
		listCmd,
1732
		psCmd,
1733
1734
		copyCmd,
		deleteCmd,
1735
		serveCmd,
1736
	} {
1737
1738
		switch cmd {
		case runCmd:
1739
1740
1741
1742
1743
			appendEnvDocs(cmd, []envconfig.EnvVar{envVars["OLLAMA_HOST"], envVars["OLLAMA_NOHISTORY"]})
		case serveCmd:
			appendEnvDocs(cmd, []envconfig.EnvVar{
				envVars["OLLAMA_DEBUG"],
				envVars["OLLAMA_HOST"],
1744
				envVars["OLLAMA_CONTEXT_LENGTH"],
1745
1746
1747
1748
1749
1750
1751
				envVars["OLLAMA_KEEP_ALIVE"],
				envVars["OLLAMA_MAX_LOADED_MODELS"],
				envVars["OLLAMA_MAX_QUEUE"],
				envVars["OLLAMA_MODELS"],
				envVars["OLLAMA_NUM_PARALLEL"],
				envVars["OLLAMA_NOPRUNE"],
				envVars["OLLAMA_ORIGINS"],
1752
				envVars["OLLAMA_SCHED_SPREAD"],
1753
				envVars["OLLAMA_FLASH_ATTENTION"],
1754
				envVars["OLLAMA_KV_CACHE_TYPE"],
1755
				envVars["OLLAMA_LLM_LIBRARY"],
1756
				envVars["OLLAMA_GPU_OVERHEAD"],
1757
				envVars["OLLAMA_LOAD_TIMEOUT"],
1758
			})
1759
1760
1761
		default:
			appendEnvDocs(cmd, envs)
		}
1762
1763
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1764
1765
	rootCmd.AddCommand(
		serveCmd,
1766
		createCmd,
Patrick Devine's avatar
Patrick Devine committed
1767
		showCmd,
1768
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1769
		stopCmd,
1770
1771
		pullCmd,
		pushCmd,
1772
1773
		signinCmd,
		signoutCmd,
Patrick Devine's avatar
Patrick Devine committed
1774
		listCmd,
1775
		psCmd,
Patrick Devine's avatar
Patrick Devine committed
1776
		copyCmd,
1777
		deleteCmd,
1778
		runnerCmd,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1779
1780
1781
1782
	)

	return rootCmd
}
1783
1784
1785
1786
1787
1788
1789
1790
1791

// If the user has explicitly set thinking options, either through the CLI or
// through the `/set think` or `set nothink` interactive options, then we
// respect them. Otherwise, we check model capabilities to see if the model
// supports thinking. If the model does support thinking, we enable it.
// Otherwise, we unset the thinking option (which is different than setting it
// to false).
//
// If capabilities are not provided, we fetch them from the server.
Michael Yang's avatar
Michael Yang committed
1792
func inferThinkingOption(caps *[]model.Capability, runOpts *runOptions, explicitlySetByUser bool) (*api.ThinkValue, error) {
1793
1794
1795
1796
1797
1798
1799
1800
1801
1802
1803
1804
1805
1806
1807
1808
1809
1810
1811
1812
1813
1814
1815
1816
1817
1818
	if explicitlySetByUser {
		return runOpts.Think, nil
	}

	if caps == nil {
		client, err := api.ClientFromEnvironment()
		if err != nil {
			return nil, err
		}
		ret, err := client.Show(context.Background(), &api.ShowRequest{
			Model: runOpts.Model,
		})
		if err != nil {
			return nil, err
		}
		caps = &ret.Capabilities
	}

	thinkingSupported := false
	for _, cap := range *caps {
		if cap == model.CapabilityThinking {
			thinkingSupported = true
		}
	}

	if thinkingSupported {
Michael Yang's avatar
Michael Yang committed
1819
		return &api.ThinkValue{Value: true}, nil
1820
1821
1822
1823
	}

	return nil, nil
}
Michael Yang's avatar
Michael Yang committed
1824
1825
1826
1827
1828
1829
1830
1831
1832
1833
1834
1835
1836
1837
1838
1839
1840
1841
1842
1843
1844
1845
1846
1847
1848
1849

func renderToolCalls(toolCalls []api.ToolCall, plainText bool) string {
	out := ""
	formatExplanation := ""
	formatValues := ""
	if !plainText {
		formatExplanation = readline.ColorGrey + readline.ColorBold
		formatValues = readline.ColorDefault
		out += formatExplanation
	}
	for i, toolCall := range toolCalls {
		argsAsJSON, err := json.Marshal(toolCall.Function.Arguments)
		if err != nil {
			return ""
		}
		if i > 0 {
			out += "\n"
		}
		// all tool calls are unexpected since we don't currently support registering any in the CLI
		out += fmt.Sprintf("  Model called a non-existent function '%s()' with arguments: %s", formatValues+toolCall.Function.Name+formatExplanation, formatValues+string(argsAsJSON)+formatExplanation)
	}
	if !plainText {
		out += readline.ColorDefault
	}
	return out
}