cmd.go 41.1 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/json"
9
	"encoding/pem"
Michael Yang's avatar
Michael Yang committed
10
	"errors"
Bruce MacDonald's avatar
Bruce MacDonald committed
11
	"fmt"
Michael Yang's avatar
Michael Yang committed
12
	"io"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
13
	"log"
14
	"math"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
15
	"net"
16
	"net/http"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
17
	"os"
18
	"os/signal"
19
	"path/filepath"
20
	"runtime"
21
	"slices"
22
	"sort"
Michael Yang's avatar
Michael Yang committed
23
	"strconv"
Michael Yang's avatar
Michael Yang committed
24
	"strings"
25
	"sync/atomic"
26
	"syscall"
Michael Yang's avatar
Michael Yang committed
27
	"time"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
28

29
	"github.com/containerd/console"
30
	"github.com/mattn/go-runewidth"
Patrick Devine's avatar
Patrick Devine committed
31
	"github.com/olekukonko/tablewriter"
Michael Yang's avatar
Michael Yang committed
32
	"github.com/spf13/cobra"
33
	"golang.org/x/crypto/ssh"
34
	"golang.org/x/sync/errgroup"
35
	"golang.org/x/term"
Michael Yang's avatar
Michael Yang committed
36

37
	"github.com/ollama/ollama/api"
38
	"github.com/ollama/ollama/envconfig"
39
	"github.com/ollama/ollama/format"
40
	"github.com/ollama/ollama/parser"
41
	"github.com/ollama/ollama/progress"
42
	"github.com/ollama/ollama/readline"
Jesse Gross's avatar
Jesse Gross committed
43
	"github.com/ollama/ollama/runner"
44
	"github.com/ollama/ollama/server"
45
	"github.com/ollama/ollama/types/model"
46
	"github.com/ollama/ollama/types/syncmap"
47
	"github.com/ollama/ollama/version"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
48
49
)

50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
// 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
	}
	for _, cap := range resp.Capabilities {
		if cap == model.CapabilityThinking {
			return
		}
	}
	fmt.Fprintf(os.Stderr, "warning: model %q does not support thinking output\n", name)
}

67
var errModelfileNotFound = errors.New("specified Modelfile wasn't found")
68
69

func getModelfileName(cmd *cobra.Command) (string, error) {
70
	filename, _ := cmd.Flags().GetString("file")
71
72
73
74
75
76

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

	absName, err := filepath.Abs(filename)
77
	if err != nil {
78
		return "", err
79
80
	}

81
	_, err = os.Stat(absName)
82
	if err != nil {
83
		return "", err
84
	}
85

86
87
88
89
	return absName, nil
}

func CreateHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
90
91
92
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

93
94
95
96
97
98
99
100
101
102
	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
103
		return err
104
105
106
107
108
109
110
111
	} else {
		f, err := os.Open(filename)
		if err != nil {
			return err
		}

		reader = f
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
112
113
	}

114
	modelfile, err := parser.ParseFile(reader)
Michael Yang's avatar
Michael Yang committed
115
116
117
118
	if err != nil {
		return err
	}

119
120
121
122
	status := "gathering model components"
	spinner := progress.NewSpinner(status)
	p.Add(status, spinner)

123
	req, err := modelfile.CreateRequest(filepath.Dir(filename))
Michael Yang's avatar
Michael Yang committed
124
125
126
	if err != nil {
		return err
	}
127
	spinner.Stop()
Michael Yang's avatar
Michael Yang committed
128

129
	req.Model = args[0]
130
131
132
133
	quantize, _ := cmd.Flags().GetString("quantize")
	if quantize != "" {
		req.Quantize = quantize
	}
134

135
136
137
138
139
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

140
141
142
143
144
145
	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 {
146
			if _, err := createBlob(cmd, client, f, digest, p); err != nil {
Michael Yang's avatar
Michael Yang committed
147
148
				return err
			}
149
150
151
152
153
154
155

			// 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
		})
156
	}
Michael Yang's avatar
Michael Yang committed
157

158
159
160
	adapters := syncmap.NewSyncMap[string, string]()
	for f, digest := range req.Adapters {
		g.Go(func() error {
161
			if _, err := createBlob(cmd, client, f, digest, p); err != nil {
Michael Yang's avatar
Michael Yang committed
162
163
				return err
			}
164
165
166
167
168

			// TODO: same here
			adapters.Store(filepath.Base(f), digest)
			return nil
		})
Michael Yang's avatar
Michael Yang committed
169
	}
Michael Yang's avatar
Michael Yang committed
170

171
172
173
174
175
176
177
	if err := g.Wait(); err != nil {
		return err
	}

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

Michael Yang's avatar
Michael Yang committed
178
	bars := make(map[string]*progress.Bar)
179
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
180
181
182
		if resp.Digest != "" {
			bar, ok := bars[resp.Digest]
			if !ok {
183
184
185
186
187
				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
188
189
190
191
192
193
194
195
196
197
198
199
200
				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)
		}

201
202
203
		return nil
	}

204
	if err := client.Create(cmd.Context(), req, fn); err != nil {
205
206
207
		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")
		}
208
209
210
211
212
213
		return err
	}

	return nil
}

214
215
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
216
217
218
219
	if err != nil {
		return "", err
	}

220
	bin, err := os.Open(realPath)
221
222
223
224
225
	if err != nil {
		return "", err
	}
	defer bin.Close()

226
227
228
229
230
231
232
233
	// Get file info to retrieve the size
	fileInfo, err := bin.Stat()
	if err != nil {
		return "", err
	}
	fileSize := fileInfo.Size()

	var pw progressWriter
234
235
236
237
	status := fmt.Sprintf("copying file %s 0%%", digest)
	spinner := progress.NewSpinner(status)
	p.Add(status, spinner)
	defer spinner.Stop()
238
239
240
241
242
243
244
245
246
247

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

	go func() {
		ticker := time.NewTicker(60 * time.Millisecond)
		defer ticker.Stop()
		for {
			select {
			case <-ticker.C:
248
				spinner.SetMessage(fmt.Sprintf("copying file %s %d%%", digest, int(100*pw.n.Load()/fileSize)))
249
			case <-done:
250
				spinner.SetMessage(fmt.Sprintf("copying file %s 100%%", digest))
251
252
253
254
255
				return
			}
		}
	}()

256
	if err := client.CreateBlob(cmd.Context(), digest, io.TeeReader(bin, &pw)); err != nil {
257
258
259
260
261
		return "", err
	}
	return digest, nil
}

262
263
264
265
266
267
268
269
270
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
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
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,
286
287
288

		// 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
289
290
291
292
293
294
295
296
297
298
299
300
301
302
	}

	return client.Generate(cmd.Context(), req, func(api.GenerateResponse) error { return nil })
}

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])
		}
303
		return err
Patrick Devine's avatar
Patrick Devine committed
304
305
306
307
	}
	return nil
}

308
func RunHandler(cmd *cobra.Command, args []string) error {
309
310
311
	interactive := true

	opts := runOptions{
312
313
		Model:    args[0],
		WordWrap: os.Getenv("TERM") == "xterm-256color",
314
		Options:  map[string]any{},
315
316
317
318
319
320
321
322
	}

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

323
324
	thinkFlag := cmd.Flags().Lookup("think")
	if thinkFlag.Changed {
Michael Yang's avatar
Michael Yang committed
325
		thinkStr, err := cmd.Flags().GetString("think")
326
327
328
		if err != nil {
			return err
		}
Michael Yang's avatar
Michael Yang committed
329
330
331
332
333
334
335
336
337
338
339
340
341

		// 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)
		}
342
343
344
345
346
347
348
349
350
	} else {
		opts.Think = nil
	}
	hidethinking, err := cmd.Flags().GetBool("hidethinking")
	if err != nil {
		return err
	}
	opts.HideThinking = hidethinking

351
352
353
354
355
356
357
358
359
360
361
362
	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}
	}

363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
	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...)
		opts.WordWrap = false
		interactive = false
	}
	opts.Prompt = strings.Join(prompts, " ")
	if len(prompts) > 0 {
		interactive = false
	}
379
380
381
382
	// Be quiet if we're redirecting to a pipe or file
	if !term.IsTerminal(int(os.Stdout.Fd())) {
		interactive = false
	}
383
384
385
386
387
388
389

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

390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
	// 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
412
413
	}

414
415
416
417
418
	opts.Think, err = inferThinkingOption(&info.Capabilities, &opts, thinkFlag.Changed)
	if err != nil {
		return err
	}

419
420
421
422
423
	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
424
425
426
427
428
429
430
431
432
433
	if len(info.ProjectorInfo) != 0 {
		opts.MultiModal = true
	}
	for k := range info.ModelInfo {
		if strings.Contains(k, ".vision.") {
			opts.MultiModal = true
			break
		}
	}

434
435
436
	opts.ParentModel = info.Details.ParentModel

	if interactive {
Patrick Devine's avatar
Patrick Devine committed
437
		if err := loadOrUnloadModel(cmd, &opts); err != nil {
Michael Yang's avatar
Michael Yang committed
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
			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()
			}
		}

453
454
455
		return generateInteractive(cmd, opts)
	}
	return generate(cmd, opts)
Bruce MacDonald's avatar
Bruce MacDonald committed
456
457
}

458
func PushHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
459
	client, err := api.ClientFromEnvironment()
460
461
462
	if err != nil {
		return err
	}
463

464
465
466
467
468
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
469
470
471
472
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)
473
474
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
475

476
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
477
		if resp.Digest != "" {
478
479
480
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
481
482
483

			bar, ok := bars[resp.Digest]
			if !ok {
484
				bar = progress.NewBar(fmt.Sprintf("pushing %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
485
486
487
488
489
490
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
491
492
493
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
494
495
496
497
498
499

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

500
501
502
		return nil
	}

Michael Yang's avatar
Michael Yang committed
503
	request := api.PushRequest{Name: args[0], Insecure: insecure}
504
505

	n := model.ParseName(args[0])
Michael Yang's avatar
Michael Yang committed
506
	if err := client.Push(cmd.Context(), &request, fn); err != nil {
507
508
509
510
511
512
		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")
		}
Michael Yang's avatar
Michael Yang committed
513
514
515
		return err
	}

516
	p.Stop()
517
	spinner.Stop()
518
519
520
521
522
523
524
525

	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
526
	return nil
527
528
}

529
func ListHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
530
	client, err := api.ClientFromEnvironment()
531
532
533
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
534

Michael Yang's avatar
Michael Yang committed
535
	models, err := client.List(cmd.Context())
Patrick Devine's avatar
Patrick Devine committed
536
537
538
539
540
541
542
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
543
		if len(args) == 0 || strings.HasPrefix(strings.ToLower(m.Name), strings.ToLower(args[0])) {
544
			data = append(data, []string{m.Name, m.Digest[:12], format.HumanBytes(m.Size), format.HumanTime(m.ModifiedAt, "Never")})
Michael Yang's avatar
Michael Yang committed
545
		}
Patrick Devine's avatar
Patrick Devine committed
546
547
548
	}

	table := tablewriter.NewWriter(os.Stdout)
Patrick Devine's avatar
Patrick Devine committed
549
	table.SetHeader([]string{"NAME", "ID", "SIZE", "MODIFIED"})
Patrick Devine's avatar
Patrick Devine committed
550
551
552
553
554
	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
555
	table.SetTablePadding("    ")
Patrick Devine's avatar
Patrick Devine committed
556
557
558
559
560
561
	table.AppendBulk(data)
	table.Render()

	return nil
}

562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
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
590
591
592
593
594
595
596
597

			var until string
			delta := time.Since(m.ExpiresAt)
			if delta > 0 {
				until = "Stopping..."
			} else {
				until = format.HumanTime(m.ExpiresAt, "Never")
			}
598
599
			ctxStr := strconv.Itoa(m.ContextLength)
			data = append(data, []string{m.Name, m.Digest[:12], format.HumanBytes(m.Size), procStr, ctxStr, until})
600
601
602
603
		}
	}

	table := tablewriter.NewWriter(os.Stdout)
604
	table.SetHeader([]string{"NAME", "ID", "SIZE", "PROCESSOR", "CONTEXT", "UNTIL"})
605
606
607
608
609
	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
610
	table.SetTablePadding("    ")
611
612
613
614
615
616
	table.AppendBulk(data)
	table.Render()

	return nil
}

617
func DeleteHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
618
	client, err := api.ClientFromEnvironment()
619
620
621
	if err != nil {
		return err
	}
622

623
624
625
626
627
628
629
630
631
632
633
	// 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 {
		if !strings.Contains(err.Error(), "not found") {
			return fmt.Errorf("unable to stop existing running model \"%s\": %s", args[0], err)
		}
	}

634
635
	for _, name := range args {
		req := api.DeleteRequest{Name: name}
Michael Yang's avatar
Michael Yang committed
636
		if err := client.Delete(cmd.Context(), &req); err != nil {
637
638
639
			return err
		}
		fmt.Printf("deleted '%s'\n", name)
640
641
642
643
	}
	return nil
}

Patrick Devine's avatar
Patrick Devine committed
644
func ShowHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
645
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
646
647
648
649
650
651
652
653
654
	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")
655
	verbose, errVerbose := cmd.Flags().GetBool("verbose")
Patrick Devine's avatar
Patrick Devine committed
656

657
	for _, boolErr := range []error{errLicense, errModelfile, errParams, errSystem, errTemplate, errVerbose} {
Patrick Devine's avatar
Patrick Devine committed
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
		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 {
692
		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
693
694
	}

695
	req := api.ShowRequest{Name: args[0], Verbose: verbose}
696
697
698
699
	resp, err := client.Show(cmd.Context(), &req)
	if err != nil {
		return err
	}
700

701
	if flagsSet == 1 {
702
703
704
705
706
707
708
709
		switch showType {
		case "license":
			fmt.Println(resp.License)
		case "modelfile":
			fmt.Println(resp.Modelfile)
		case "parameters":
			fmt.Println(resp.Parameters)
		case "system":
710
			fmt.Print(resp.System)
711
		case "template":
712
			fmt.Print(resp.Template)
713
714
715
		}

		return nil
Patrick Devine's avatar
Patrick Devine committed
716
717
	}

718
	return showInfo(resp, verbose, os.Stdout)
719
720
}

721
func showInfo(resp *api.ShowResponse, verbose bool, w io.Writer) error {
Michael Yang's avatar
Michael Yang committed
722
723
724
725
726
727
728
	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("    ")
729

Michael Yang's avatar
Michael Yang committed
730
731
732
		switch header {
		case "Template", "System", "License":
			table.SetColWidth(100)
733
734
		}

Michael Yang's avatar
Michael Yang committed
735
736
737
		table.AppendBulk(rows())
		table.Render()
		fmt.Fprintln(w)
Patrick Devine's avatar
Patrick Devine committed
738
739
	}

Michael Yang's avatar
Michael Yang committed
740
741
742
743
744
745
746
747
748
749
750
751
752
753
	tableRender("Model", func() (rows [][]string) {
		if resp.ModelInfo != nil {
			arch := resp.ModelInfo["general.architecture"].(string)
			rows = append(rows, []string{"", "architecture", arch})
			rows = append(rows, []string{"", "parameters", format.HumanNumber(uint64(resp.ModelInfo["general.parameter_count"].(float64)))})
			rows = append(rows, []string{"", "context length", strconv.FormatFloat(resp.ModelInfo[fmt.Sprintf("%s.context_length", arch)].(float64), 'f', -1, 64)})
			rows = append(rows, []string{"", "embedding length", strconv.FormatFloat(resp.ModelInfo[fmt.Sprintf("%s.embedding_length", arch)].(float64), 'f', -1, 64)})
		} 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
	})
754

755
756
757
758
759
760
761
762
763
	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
764
765
766
767
768
769
770
771
772
	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
		})
773
774
	}

Michael Yang's avatar
Michael Yang committed
775
776
777
778
779
780
781
782
783
784
	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
		})
785
786
	}

787
788
789
790
791
792
793
794
795
796
797
	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) {
798
799
				case bool:
					v = fmt.Sprintf("%t", vData)
800
801
802
803
804
				case string:
					v = vData
				case float64:
					v = fmt.Sprintf("%g", vData)
				case []any:
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
					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)
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
					}
				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
855
856
	head := func(s string, n int) (rows [][]string) {
		scanner := bufio.NewScanner(strings.NewReader(s))
857
858
859
860
861
		count := 0
		for scanner.Scan() {
			text := strings.TrimSpace(scanner.Text())
			if text == "" {
				continue
862
			}
863
864
865
866
867
868
869
			count++
			if n < 0 || count <= n {
				rows = append(rows, []string{"", text})
			}
		}
		if n >= 0 && count > n {
			rows = append(rows, []string{"", "..."})
870
		}
Michael Yang's avatar
Michael Yang committed
871
		return
872
873
	}

Michael Yang's avatar
Michael Yang committed
874
875
876
877
878
	if resp.System != "" {
		tableRender("System", func() [][]string {
			return head(resp.System, 2)
		})
	}
879

Michael Yang's avatar
Michael Yang committed
880
881
882
883
	if resp.License != "" {
		tableRender("License", func() [][]string {
			return head(resp.License, 2)
		})
884
	}
Michael Yang's avatar
Michael Yang committed
885
886

	return nil
887
888
}

Patrick Devine's avatar
Patrick Devine committed
889
func CopyHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
890
	client, err := api.ClientFromEnvironment()
891
892
893
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
894
895

	req := api.CopyRequest{Source: args[0], Destination: args[1]}
Michael Yang's avatar
Michael Yang committed
896
	if err := client.Copy(cmd.Context(), &req); err != nil {
Patrick Devine's avatar
Patrick Devine committed
897
898
899
900
901
902
		return err
	}
	fmt.Printf("copied '%s' to '%s'\n", args[0], args[1])
	return nil
}

903
func PullHandler(cmd *cobra.Command, args []string) error {
904
905
906
907
908
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
909
	client, err := api.ClientFromEnvironment()
910
911
912
	if err != nil {
		return err
	}
913

Michael Yang's avatar
Michael Yang committed
914
915
916
917
918
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

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

919
920
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
921

922
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
923
		if resp.Digest != "" {
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
			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
			}

944
945
946
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
947
948
949

			bar, ok := bars[resp.Digest]
			if !ok {
950
951
952
953
954
955
				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
956
957
958
959
960
961
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
962
963
964
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
965
966
967
968
969
970

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

971
972
		return nil
	}
973

Michael Yang's avatar
Michael Yang committed
974
	request := api.PullRequest{Name: args[0], Insecure: insecure}
975
	return client.Pull(cmd.Context(), &request, fn)
Michael Yang's avatar
Michael Yang committed
976
977
}

978
979
type generateContextKey string

980
type runOptions struct {
981
982
983
984
985
986
987
988
989
990
991
	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
992
	Think        *api.ThinkValue
993
	HideThinking bool
994
995
}

996
997
998
999
1000
1001
1002
1003
1004
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
1005
1006
			if state.lineLength+1 > termWidth-5 {
				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
1007
1008
1009
1010
1011
1012
1013
					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
1014
1015
				a := runewidth.StringWidth(state.wordBuffer)
				if a > 0 {
1016
					fmt.Printf("\x1b[%dD", a)
1017
1018
				}
				fmt.Printf("\x1b[K\n")
1019
				fmt.Printf("%s%c", state.wordBuffer, ch)
1020
1021
1022
				chWidth := runewidth.RuneWidth(ch)

				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
1023
1024
			} else {
				fmt.Print(string(ch))
1025
1026
1027
1028
				state.lineLength += runewidth.RuneWidth(ch)
				if runewidth.RuneWidth(ch) >= 2 {
					state.wordBuffer = ""
					continue
Josh Yan's avatar
Josh Yan committed
1029
				}
1030
1031

				switch ch {
Michael Yang's avatar
Michael Yang committed
1032
				case ' ', '\t':
1033
					state.wordBuffer = ""
Michael Yang's avatar
Michael Yang committed
1034
				case '\n', '\r':
1035
					state.lineLength = 0
Michael Yang's avatar
Michael Yang committed
1036
					state.wordBuffer = ""
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
				default:
					state.wordBuffer += string(ch)
				}
			}
		}
	} else {
		fmt.Printf("%s%s", state.wordBuffer, content)
		if len(state.wordBuffer) > 0 {
			state.wordBuffer = ""
		}
	}
}

1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
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
}

1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
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
1094
	var thinkingContent strings.Builder
1095
1096
	var latest api.ChatResponse
	var fullResponse strings.Builder
1097
1098
	var thinkTagOpened bool = false
	var thinkTagClosed bool = false
1099

1100
1101
	role := "assistant"

1102
	fn := func(response api.ChatResponse) error {
1103
1104
1105
		if response.Message.Content != "" || !opts.HideThinking {
			p.StopAndClear()
		}
1106
1107
1108
1109

		latest = response

		role = response.Message.Role
1110
1111
1112
1113
		if response.Message.Thinking != "" && !opts.HideThinking {
			if !thinkTagOpened {
				fmt.Print(thinkingOutputOpeningText(false))
				thinkTagOpened = true
Michael Yang's avatar
Michael Yang committed
1114
				thinkTagClosed = false
1115
			}
Michael Yang's avatar
Michael Yang committed
1116
			thinkingContent.WriteString(response.Message.Thinking)
1117
1118
1119
			displayResponse(response.Message.Thinking, opts.WordWrap, state)
		}

1120
		content := response.Message.Content
Michael Yang's avatar
Michael Yang committed
1121
1122
1123
1124
		if thinkTagOpened && !thinkTagClosed && (content != "" || len(response.Message.ToolCalls) > 0) {
			if !strings.HasSuffix(thinkingContent.String(), "\n") {
				fmt.Println()
			}
1125
			fmt.Print(thinkingOutputClosingText(false))
Michael Yang's avatar
Michael Yang committed
1126
			thinkTagOpened = false
1127
			thinkTagClosed = true
Michael Yang's avatar
Michael Yang committed
1128
			state = &displayResponseState{}
1129
1130
1131
1132
1133
		}
		// 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)
1134
1135
		fullResponse.WriteString(content)

Michael Yang's avatar
Michael Yang committed
1136
1137
1138
1139
1140
1141
1142
		if response.Message.ToolCalls != nil {
			toolCalls := response.Message.ToolCalls
			if len(toolCalls) > 0 {
				fmt.Print(renderToolCalls(toolCalls, false))
			}
		}

1143
1144
1145
1146
1147
		displayResponse(content, opts.WordWrap, state)

		return nil
	}

1148
1149
1150
1151
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

1152
1153
1154
	req := &api.ChatRequest{
		Model:    opts.Model,
		Messages: opts.Messages,
1155
		Format:   json.RawMessage(opts.Format),
1156
		Options:  opts.Options,
1157
		Think:    opts.Think,
1158
1159
	}

1160
1161
1162
1163
	if opts.KeepAlive != nil {
		req.KeepAlive = opts.KeepAlive
	}

1164
1165
1166
1167
	if err := client.Chat(cancelCtx, req, fn); err != nil {
		if errors.Is(err, context.Canceled) {
			return nil, nil
		}
1168
1169
1170
1171
1172
1173
1174
1175

		// 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
		}
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
		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
1197
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
1198
	if err != nil {
1199
		return err
Patrick Devine's avatar
Patrick Devine committed
1200
	}
Michael Yang's avatar
Michael Yang committed
1201

Michael Yang's avatar
Michael Yang committed
1202
	p := progress.NewProgress(os.Stderr)
1203
	defer p.StopAndClear()
1204

Michael Yang's avatar
Michael Yang committed
1205
1206
1207
	spinner := progress.NewSpinner("")
	p.Add("", spinner)

1208
1209
1210
1211
1212
1213
1214
	var latest api.GenerateResponse

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

Michael Yang's avatar
Michael Yang committed
1215
	ctx, cancel := context.WithCancel(cmd.Context())
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
	defer cancel()

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

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

1226
	var state *displayResponseState = &displayResponseState{}
Michael Yang's avatar
Michael Yang committed
1227
	var thinkingContent strings.Builder
1228
1229
	var thinkTagOpened bool = false
	var thinkTagClosed bool = false
1230

1231
	plainText := !term.IsTerminal(int(os.Stdout.Fd()))
1232

1233
	fn := func(response api.GenerateResponse) error {
Patrick Devine's avatar
Patrick Devine committed
1234
		latest = response
1235
		content := response.Response
1236

1237
1238
1239
1240
1241
1242
1243
1244
		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
1245
				thinkTagClosed = false
1246
			}
Michael Yang's avatar
Michael Yang committed
1247
			thinkingContent.WriteString(response.Thinking)
1248
1249
1250
			displayResponse(response.Thinking, opts.WordWrap, state)
		}

Michael Yang's avatar
Michael Yang committed
1251
1252
1253
1254
		if thinkTagOpened && !thinkTagClosed && (content != "" || len(response.ToolCalls) > 0) {
			if !strings.HasSuffix(thinkingContent.String(), "\n") {
				fmt.Println()
			}
1255
			fmt.Print(thinkingOutputClosingText(plainText))
Michael Yang's avatar
Michael Yang committed
1256
			thinkTagOpened = false
1257
			thinkTagClosed = true
Michael Yang's avatar
Michael Yang committed
1258
			state = &displayResponseState{}
1259
1260
		}

1261
		displayResponse(content, opts.WordWrap, state)
1262

Michael Yang's avatar
Michael Yang committed
1263
1264
1265
1266
1267
1268
1269
		if response.ToolCalls != nil {
			toolCalls := response.ToolCalls
			if len(toolCalls) > 0 {
				fmt.Print(renderToolCalls(toolCalls, plainText))
			}
		}

Patrick Devine's avatar
Patrick Devine committed
1270
1271
		return nil
	}
1272

1273
1274
1275
1276
1277
1278
1279
	if opts.MultiModal {
		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
		if err != nil {
			return err
		}
	}

1280
1281
1282
1283
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

Michael Yang's avatar
Michael Yang committed
1284
	request := api.GenerateRequest{
1285
1286
1287
1288
		Model:     opts.Model,
		Prompt:    opts.Prompt,
		Context:   generateContext,
		Images:    opts.Images,
1289
		Format:    json.RawMessage(opts.Format),
1290
1291
1292
		System:    opts.System,
		Options:   opts.Options,
		KeepAlive: opts.KeepAlive,
1293
		Think:     opts.Think,
Michael Yang's avatar
Michael Yang committed
1294
1295
1296
	}

	if err := client.Generate(ctx, &request, fn); err != nil {
1297
		if errors.Is(err, context.Canceled) {
1298
			return nil
1299
		}
1300
		return err
Patrick Devine's avatar
Patrick Devine committed
1301
	}
1302

1303
	if opts.Prompt != "" {
Michael Yang's avatar
Michael Yang committed
1304
1305
		fmt.Println()
		fmt.Println()
Patrick Devine's avatar
Patrick Devine committed
1306
	}
1307

1308
1309
1310
1311
	if !latest.Done {
		return nil
	}

Patrick Devine's avatar
Patrick Devine committed
1312
1313
	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
1314
		return err
Patrick Devine's avatar
Patrick Devine committed
1315
	}
Michael Yang's avatar
Michael Yang committed
1316

Patrick Devine's avatar
Patrick Devine committed
1317
1318
	if verbose {
		latest.Summary()
Michael Yang's avatar
Michael Yang committed
1319
	}
Michael Yang's avatar
Michael Yang committed
1320

Patrick Devine's avatar
Patrick Devine committed
1321
1322
1323
	ctx = context.WithValue(cmd.Context(), generateContextKey("context"), latest.Context)
	cmd.SetContext(ctx)

1324
	return nil
Michael Yang's avatar
Michael Yang committed
1325
1326
}

1327
func RunServer(_ *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1328
	if err := initializeKeypair(); err != nil {
1329
1330
1331
		return err
	}

Michael Yang's avatar
host  
Michael Yang committed
1332
	ln, err := net.Listen("tcp", envconfig.Host().Host)
1333
1334
1335
	if err != nil {
		return err
	}
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1336

1337
1338
1339
1340
1341
1342
	err = server.Serve(ln)
	if errors.Is(err, http.ErrServerClosed) {
		return nil
	}

	return err
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1343
1344
}

1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
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
1357
		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
1358
1359
1360
1361
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1362
		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
1363
1364
1365
1366
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1367
		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
1368
1369
1370
			return fmt.Errorf("could not create directory %w", err)
		}

Michael Yang's avatar
Michael Yang committed
1371
		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
1372
1373
1374
			return err
		}

Michael Yang's avatar
Michael Yang committed
1375
		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
1376
1377
1378
1379
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1380
		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
1381

Michael Yang's avatar
Michael Yang committed
1382
		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
1383
1384
1385
			return err
		}

Michael Yang's avatar
Michael Yang committed
1386
		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
1387
1388
1389
1390
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1391
func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1392
	client, err := api.ClientFromEnvironment()
1393
1394
1395
	if err != nil {
		return err
	}
Michael Yang's avatar
Michael Yang committed
1396
	if err := client.Heartbeat(cmd.Context()); err != nil {
1397
		if !(strings.Contains(err.Error(), " refused") || strings.Contains(err.Error(), "could not connect")) {
Bruce MacDonald's avatar
Bruce MacDonald committed
1398
1399
			return err
		}
1400
		if err := startApp(cmd.Context(), client); err != nil {
1401
			return fmt.Errorf("ollama server not responding - %w", err)
1402
1403
1404
1405
1406
		}
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1407
1408
1409
1410
1411
1412
1413
1414
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
1415
1416
1417
1418
1419
		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
1420
1421
	}

1422
	if serverVersion != version.Version {
Michael Yang's avatar
Michael Yang committed
1423
		fmt.Printf("Warning: client version is %s\n", version.Version)
1424
	}
Michael Yang's avatar
Michael Yang committed
1425
1426
}

1427
func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
1428
1429
1430
1431
1432
	if len(envs) == 0 {
		return
	}

	envUsage := `
1433
1434
Environment Variables:
`
1435
	for _, e := range envs {
1436
		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
1437
1438
1439
	}

	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
1440
1441
}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1442
1443
func NewCLI() *cobra.Command {
	log.SetFlags(log.LstdFlags | log.Lshortfile)
Michael Yang's avatar
Michael Yang committed
1444
	cobra.EnableCommandSorting = false
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1445

1446
	if runtime.GOOS == "windows" && term.IsTerminal(int(os.Stdout.Fd())) {
1447
		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
1448
1449
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1450
	rootCmd := &cobra.Command{
1451
1452
1453
1454
		Use:           "ollama",
		Short:         "Large language model runner",
		SilenceUsage:  true,
		SilenceErrors: true,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1455
1456
1457
		CompletionOptions: cobra.CompletionOptions{
			DisableDefaultCmd: true,
		},
Michael Yang's avatar
Michael Yang committed
1458
1459
1460
1461
1462
1463
1464
1465
		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
1466
1467
	}

Michael Yang's avatar
Michael Yang committed
1468
	rootCmd.Flags().BoolP("version", "v", false, "Show version information")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1469

1470
	createCmd := &cobra.Command{
1471
		Use:     "create MODEL",
1472
		Short:   "Create a model",
Michael Yang's avatar
Michael Yang committed
1473
		Args:    cobra.ExactArgs(1),
1474
1475
		PreRunE: checkServerHeartbeat,
		RunE:    CreateHandler,
1476
1477
	}

1478
	createCmd.Flags().StringP("file", "f", "", "Name of the Modelfile (default \"Modelfile\")")
1479
	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_K_M)")
1480

Patrick Devine's avatar
Patrick Devine committed
1481
1482
1483
	showCmd := &cobra.Command{
		Use:     "show MODEL",
		Short:   "Show information for a model",
Michael Yang's avatar
Michael Yang committed
1484
		Args:    cobra.ExactArgs(1),
Patrick Devine's avatar
Patrick Devine committed
1485
1486
1487
1488
1489
1490
1491
1492
		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")
1493
	showCmd.Flags().Bool("system", false, "Show system message of a model")
1494
	showCmd.Flags().BoolP("verbose", "v", false, "Show detailed model information")
Patrick Devine's avatar
Patrick Devine committed
1495

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1496
	runCmd := &cobra.Command{
1497
1498
1499
1500
1501
		Use:     "run MODEL [PROMPT]",
		Short:   "Run a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    RunHandler,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1502
1503
	}

1504
	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
1505
	runCmd.Flags().Bool("verbose", false, "Show timings for response")
1506
	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
1507
	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1508
	runCmd.Flags().String("format", "", "Response format (e.g. json)")
Michael Yang's avatar
Michael Yang committed
1509
1510
	runCmd.Flags().String("think", "", "Enable thinking mode: true/false or high/medium/low for supported models")
	runCmd.Flags().Lookup("think").NoOptDefVal = "true"
1511
	runCmd.Flags().Bool("hidethinking", false, "Hide thinking output (if provided)")
Patrick Devine's avatar
Patrick Devine committed
1512
1513
1514
1515
1516
1517
1518
1519
1520

	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
1521
1522
1523
1524
	serveCmd := &cobra.Command{
		Use:     "serve",
		Aliases: []string{"start"},
		Short:   "Start ollama",
Michael Yang's avatar
Michael Yang committed
1525
		Args:    cobra.ExactArgs(0),
Michael Yang's avatar
Michael Yang committed
1526
		RunE:    RunServer,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1527
1528
	}

1529
	pullCmd := &cobra.Command{
1530
1531
		Use:     "pull MODEL",
		Short:   "Pull a model from a registry",
Michael Yang's avatar
Michael Yang committed
1532
		Args:    cobra.ExactArgs(1),
1533
1534
		PreRunE: checkServerHeartbeat,
		RunE:    PullHandler,
1535
1536
	}

1537
1538
	pullCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1539
	pushCmd := &cobra.Command{
1540
1541
		Use:     "push MODEL",
		Short:   "Push a model to a registry",
Michael Yang's avatar
Michael Yang committed
1542
		Args:    cobra.ExactArgs(1),
1543
1544
		PreRunE: checkServerHeartbeat,
		RunE:    PushHandler,
1545
1546
	}

1547
1548
	pushCmd.Flags().Bool("insecure", false, "Use an insecure registry")

Patrick Devine's avatar
Patrick Devine committed
1549
	listCmd := &cobra.Command{
1550
		Use:     "list",
Patrick Devine's avatar
Patrick Devine committed
1551
		Aliases: []string{"ls"},
1552
		Short:   "List models",
1553
		PreRunE: checkServerHeartbeat,
1554
		RunE:    ListHandler,
1555
	}
1556
1557
1558
1559
1560
1561
1562

	psCmd := &cobra.Command{
		Use:     "ps",
		Short:   "List running models",
		PreRunE: checkServerHeartbeat,
		RunE:    ListRunningHandler,
	}
Patrick Devine's avatar
Patrick Devine committed
1563
	copyCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1564
		Use:     "cp SOURCE DESTINATION",
1565
		Short:   "Copy a model",
Michael Yang's avatar
Michael Yang committed
1566
		Args:    cobra.ExactArgs(2),
1567
1568
		PreRunE: checkServerHeartbeat,
		RunE:    CopyHandler,
Patrick Devine's avatar
Patrick Devine committed
1569
1570
	}

1571
	deleteCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1572
		Use:     "rm MODEL [MODEL...]",
1573
1574
1575
1576
		Short:   "Remove a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    DeleteHandler,
Patrick Devine's avatar
Patrick Devine committed
1577
1578
	}

1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
	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:])
	})

1591
1592
1593
	envVars := envconfig.AsMap()

	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
1594

1595
1596
1597
1598
	for _, cmd := range []*cobra.Command{
		createCmd,
		showCmd,
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1599
		stopCmd,
1600
1601
1602
		pullCmd,
		pushCmd,
		listCmd,
1603
		psCmd,
1604
1605
		copyCmd,
		deleteCmd,
1606
		serveCmd,
1607
	} {
1608
1609
		switch cmd {
		case runCmd:
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
			appendEnvDocs(cmd, []envconfig.EnvVar{envVars["OLLAMA_HOST"], envVars["OLLAMA_NOHISTORY"]})
		case serveCmd:
			appendEnvDocs(cmd, []envconfig.EnvVar{
				envVars["OLLAMA_DEBUG"],
				envVars["OLLAMA_HOST"],
				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"],
1622
				envVars["OLLAMA_SCHED_SPREAD"],
1623
				envVars["OLLAMA_FLASH_ATTENTION"],
1624
				envVars["OLLAMA_KV_CACHE_TYPE"],
1625
				envVars["OLLAMA_LLM_LIBRARY"],
1626
				envVars["OLLAMA_GPU_OVERHEAD"],
1627
				envVars["OLLAMA_LOAD_TIMEOUT"],
1628
			})
1629
1630
1631
		default:
			appendEnvDocs(cmd, envs)
		}
1632
1633
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1634
1635
	rootCmd.AddCommand(
		serveCmd,
1636
		createCmd,
Patrick Devine's avatar
Patrick Devine committed
1637
		showCmd,
1638
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1639
		stopCmd,
1640
1641
		pullCmd,
		pushCmd,
Patrick Devine's avatar
Patrick Devine committed
1642
		listCmd,
1643
		psCmd,
Patrick Devine's avatar
Patrick Devine committed
1644
		copyCmd,
1645
		deleteCmd,
1646
		runnerCmd,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1647
1648
1649
1650
	)

	return rootCmd
}
1651
1652
1653
1654
1655
1656
1657
1658
1659

// 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
1660
func inferThinkingOption(caps *[]model.Capability, runOpts *runOptions, explicitlySetByUser bool) (*api.ThinkValue, error) {
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
	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
1687
		return &api.ThinkValue{Value: true}, nil
1688
1689
1690
1691
	}

	return nil, nil
}
Michael Yang's avatar
Michael Yang committed
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
1707
1708
1709
1710
1711
1712
1713
1714
1715
1716
1717

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
}