cmd.go 33.6 KB
Newer Older
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1
2
3
package cmd

import (
4
	"archive/zip"
Michael Yang's avatar
Michael Yang committed
5
	"bufio"
Michael Yang's avatar
Michael Yang committed
6
	"bytes"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
7
	"context"
8
9
	"crypto/ed25519"
	"crypto/rand"
Michael Yang's avatar
Michael Yang committed
10
	"crypto/sha256"
11
	"encoding/pem"
Michael Yang's avatar
Michael Yang committed
12
	"errors"
Bruce MacDonald's avatar
Bruce MacDonald committed
13
	"fmt"
Michael Yang's avatar
Michael Yang committed
14
	"io"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
15
	"log"
16
	"math"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
17
	"net"
18
	"net/http"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
19
	"os"
20
	"os/signal"
21
	"path/filepath"
22
	"runtime"
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/term"
Michael Yang's avatar
Michael Yang committed
35

36
	"github.com/ollama/ollama/api"
37
	"github.com/ollama/ollama/envconfig"
38
	"github.com/ollama/ollama/format"
39
	"github.com/ollama/ollama/parser"
40
41
42
	"github.com/ollama/ollama/progress"
	"github.com/ollama/ollama/server"
	"github.com/ollama/ollama/version"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
43
44
)

45
46
47
48
49
50
51
52
53
54
55
56
57
58
var (
	errModelNotFound     = errors.New("no Modelfile or safetensors files found")
	errModelfileNotFound = errors.New("specified Modelfile wasn't found")
)

func getModelfileName(cmd *cobra.Command) (string, error) {
	fn, _ := cmd.Flags().GetString("file")

	filename := fn
	if filename == "" {
		filename = "Modelfile"
	}

	absName, err := filepath.Abs(filename)
59
	if err != nil {
60
		return "", err
61
62
	}

63
	_, err = os.Stat(absName)
64
	if err != nil {
65
		return fn, err
66
	}
67

68
69
70
71
	return absName, nil
}

func CreateHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
72
73
74
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

75
76
77
78
79
80
81
82
83
84
	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
85
		return err
86
87
88
89
90
91
92
93
	} else {
		f, err := os.Open(filename)
		if err != nil {
			return err
		}

		reader = f
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
94
95
	}

96
	modelfile, err := parser.ParseFile(reader)
Michael Yang's avatar
Michael Yang committed
97
98
99
100
101
102
103
104
105
	if err != nil {
		return err
	}

	home, err := os.UserHomeDir()
	if err != nil {
		return err
	}

106
107
	status := "transferring model data"
	spinner := progress.NewSpinner(status)
108
	p.Add(status, spinner)
109
	defer p.Stop()
110

111
112
113
114
115
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
116
117
	for i := range modelfile.Commands {
		switch modelfile.Commands[i].Name {
Michael Yang's avatar
Michael Yang committed
118
		case "model", "adapter":
Michael Yang's avatar
Michael Yang committed
119
			path := modelfile.Commands[i].Args
Michael Yang's avatar
Michael Yang committed
120
121
122
123
124
125
			if path == "~" {
				path = home
			} else if strings.HasPrefix(path, "~/") {
				path = filepath.Join(home, path[2:])
			}

126
127
128
129
			if !filepath.IsAbs(path) {
				path = filepath.Join(filepath.Dir(filename), path)
			}

130
			fi, err := os.Stat(path)
Michael Yang's avatar
Michael Yang committed
131
			if errors.Is(err, os.ErrNotExist) && modelfile.Commands[i].Name == "model" {
Michael Yang's avatar
Michael Yang committed
132
				continue
Michael Yang's avatar
Michael Yang committed
133
134
135
136
			} else if err != nil {
				return err
			}

137
			if fi.IsDir() {
Michael Yang's avatar
Michael Yang committed
138
139
140
				// this is likely a safetensors or pytorch directory
				// TODO make this work w/ adapters
				tempfile, err := tempZipFiles(path)
141
142
143
				if err != nil {
					return err
				}
Michael Yang's avatar
Michael Yang committed
144
				defer os.RemoveAll(tempfile)
145

Michael Yang's avatar
Michael Yang committed
146
				path = tempfile
Michael Yang's avatar
Michael Yang committed
147
148
			}

149
			digest, err := createBlob(cmd, client, path, spinner)
150
			if err != nil {
Michael Yang's avatar
Michael Yang committed
151
152
153
				return err
			}

Michael Yang's avatar
Michael Yang committed
154
			modelfile.Commands[i].Args = "@" + digest
Michael Yang's avatar
Michael Yang committed
155
156
		}
	}
Michael Yang's avatar
Michael Yang committed
157

Michael Yang's avatar
Michael Yang committed
158
	bars := make(map[string]*progress.Bar)
159
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
160
161
162
163
164
		if resp.Digest != "" {
			spinner.Stop()

			bar, ok := bars[resp.Digest]
			if !ok {
165
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
166
167
168
169
170
171
172
173
174
175
176
177
178
				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)
		}

179
180
181
		return nil
	}

182
	quantize, _ := cmd.Flags().GetString("quantize")
Michael Yang's avatar
Michael Yang committed
183

184
	request := api.CreateRequest{Name: args[0], Modelfile: modelfile.String(), Quantize: quantize}
Michael Yang's avatar
Michael Yang committed
185
	if err := client.Create(cmd.Context(), &request, fn); err != nil {
186
187
188
189
190
191
		return err
	}

	return nil
}

Michael Yang's avatar
Michael Yang committed
192
193
194
195
196
197
198
func tempZipFiles(path string) (string, error) {
	tempfile, err := os.CreateTemp("", "ollama-tf")
	if err != nil {
		return "", err
	}
	defer tempfile.Close()

Michael Yang's avatar
Michael Yang committed
199
200
	detectContentType := func(path string) (string, error) {
		f, err := os.Open(path)
Michael Yang's avatar
Michael Yang committed
201
202
203
		if err != nil {
			return "", err
		}
Michael Yang's avatar
Michael Yang committed
204
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
205

Michael Yang's avatar
Michael Yang committed
206
207
		var b bytes.Buffer
		b.Grow(512)
Michael Yang's avatar
Michael Yang committed
208

Michael Yang's avatar
Michael Yang committed
209
210
211
212
213
214
		if _, err := io.CopyN(&b, f, 512); err != nil && !errors.Is(err, io.EOF) {
			return "", err
		}

		contentType, _, _ := strings.Cut(http.DetectContentType(b.Bytes()), ";")
		return contentType, nil
Michael Yang's avatar
Michael Yang committed
215
216
	}

Michael Yang's avatar
Michael Yang committed
217
218
219
220
221
222
223
224
225
226
227
	glob := func(pattern, contentType string) ([]string, error) {
		matches, err := filepath.Glob(pattern)
		if err != nil {
			return nil, err
		}

		for _, safetensor := range matches {
			if ct, err := detectContentType(safetensor); err != nil {
				return nil, err
			} else if ct != contentType {
				return nil, fmt.Errorf("invalid content type: expected %s for %s", ct, safetensor)
Michael Yang's avatar
Michael Yang committed
228
			}
Michael Yang's avatar
Michael Yang committed
229
230
231
232
233
234
235
236
237
238
		}

		return matches, nil
	}

	var files []string
	if st, _ := glob(filepath.Join(path, "model*.safetensors"), "application/octet-stream"); len(st) > 0 {
		// safetensors files might be unresolved git lfs references; skip if they are
		// covers model-x-of-y.safetensors, model.fp32-x-of-y.safetensors, model.safetensors
		files = append(files, st...)
239
240
241
242
243
244
	} else if st, _ := glob(filepath.Join(path, "adapters.safetensors"), "application/octet-stream"); len(st) > 0 {
		// covers adapters.safetensors
		files = append(files, st...)
	} else if st, _ := glob(filepath.Join(path, "adapter_model.safetensors"), "application/octet-stream"); len(st) > 0 {
		// covers adapter_model.safetensors
		files = append(files, st...)
Michael Yang's avatar
Michael Yang committed
245
246
247
248
	} else if pt, _ := glob(filepath.Join(path, "pytorch_model*.bin"), "application/zip"); len(pt) > 0 {
		// pytorch files might also be unresolved git lfs references; skip if they are
		// covers pytorch_model-x-of-y.bin, pytorch_model.fp32-x-of-y.bin, pytorch_model.bin
		files = append(files, pt...)
Patrick Devine's avatar
Patrick Devine committed
249
	} else if pt, _ := glob(filepath.Join(path, "consolidated*.pth"), "application/zip"); len(pt) > 0 {
Michael Yang's avatar
Michael Yang committed
250
251
252
253
		// pytorch files might also be unresolved git lfs references; skip if they are
		// covers consolidated.x.pth, consolidated.pth
		files = append(files, pt...)
	} else {
254
		return "", errModelNotFound
Michael Yang's avatar
Michael Yang committed
255
256
257
258
259
260
261
262
263
	}

	// add configuration files, json files are detected as text/plain
	js, err := glob(filepath.Join(path, "*.json"), "text/plain")
	if err != nil {
		return "", err
	}
	files = append(files, js...)

Michael Yang's avatar
Michael Yang committed
264
265
266
267
268
269
270
271
	// bert models require a nested config.json
	// TODO(mxyng): merge this with the glob above
	js, err = glob(filepath.Join(path, "**/*.json"), "text/plain")
	if err != nil {
		return "", err
	}
	files = append(files, js...)

Michael Yang's avatar
Michael Yang committed
272
273
274
275
276
277
278
279
280
	if tks, _ := glob(filepath.Join(path, "tokenizer.model"), "application/octet-stream"); len(tks) > 0 {
		// add tokenizer.model if it exists, tokenizer.json is automatically picked up by the previous glob
		// tokenizer.model might be a unresolved git lfs reference; error if it is
		files = append(files, tks...)
	} else if tks, _ := glob(filepath.Join(path, "**/tokenizer.model"), "text/plain"); len(tks) > 0 {
		// some times tokenizer.model is in a subdirectory (e.g. meta-llama/Meta-Llama-3-8B)
		files = append(files, tks...)
	}

281
282
283
	zipfile := zip.NewWriter(tempfile)
	defer zipfile.Close()

Michael Yang's avatar
Michael Yang committed
284
285
286
	for _, file := range files {
		f, err := os.Open(file)
		if err != nil {
Michael Yang's avatar
Michael Yang committed
287
288
			return "", err
		}
Michael Yang's avatar
Michael Yang committed
289
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
290
291
292
293
294
295

		fi, err := f.Stat()
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
296
		zfi, err := zip.FileInfoHeader(fi)
Michael Yang's avatar
Michael Yang committed
297
298
299
300
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
301
302
303
304
305
		zfi.Name, err = filepath.Rel(path, file)
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
306
		zf, err := zipfile.CreateHeader(zfi)
Michael Yang's avatar
Michael Yang committed
307
308
309
310
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
311
		if _, err := io.Copy(zf, f); err != nil {
Michael Yang's avatar
Michael Yang committed
312
313
314
315
316
317
318
			return "", err
		}
	}

	return tempfile.Name(), nil
}

319
func createBlob(cmd *cobra.Command, client *api.Client, path string, spinner *progress.Spinner) (string, error) {
320
321
322
323
324
325
	bin, err := os.Open(path)
	if err != nil {
		return "", err
	}
	defer bin.Close()

326
327
328
329
330
331
332
	// Get file info to retrieve the size
	fileInfo, err := bin.Stat()
	if err != nil {
		return "", err
	}
	fileSize := fileInfo.Size()

333
334
335
336
	hash := sha256.New()
	if _, err := io.Copy(hash, bin); err != nil {
		return "", err
	}
337
338
339
340

	if _, err := bin.Seek(0, io.SeekStart); err != nil {
		return "", err
	}
341

342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
	var pw progressWriter
	status := "transferring model data 0%"
	spinner.SetMessage(status)

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

	go func() {
		ticker := time.NewTicker(60 * time.Millisecond)
		defer ticker.Stop()
		for {
			select {
			case <-ticker.C:
				spinner.SetMessage(fmt.Sprintf("transferring model data %d%%", int(100*pw.n.Load()/fileSize)))
			case <-done:
				spinner.SetMessage("transferring model data 100%")
				return
			}
		}
	}()

363
	digest := fmt.Sprintf("sha256:%x", hash.Sum(nil))
364
	if err = client.CreateBlob(cmd.Context(), digest, io.TeeReader(bin, &pw)); err != nil {
365
366
367
368
369
		return "", err
	}
	return digest, nil
}

370
371
372
373
374
375
376
377
378
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
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
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,
	}

	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])
		}
	}
	return nil
}

412
func RunHandler(cmd *cobra.Command, args []string) error {
413
414
415
	interactive := true

	opts := runOptions{
416
417
418
		Model:    args[0],
		WordWrap: os.Getenv("TERM") == "xterm-256color",
		Options:  map[string]interface{}{},
419
420
421
422
423
424
425
426
	}

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

427
428
429
430
431
432
433
434
435
436
437
438
	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}
	}

439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
	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
	}
455
456
457
458
	// Be quiet if we're redirecting to a pipe or file
	if !term.IsTerminal(int(os.Stdout.Fd())) {
		interactive = false
	}
459
460
461
462
463
464
465

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

466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
	// 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
488
489
	}

490
	opts.MultiModal = len(info.ProjectorInfo) != 0
491
492
493
	opts.ParentModel = info.Details.ParentModel

	if interactive {
Patrick Devine's avatar
Patrick Devine committed
494
		if err := loadOrUnloadModel(cmd, &opts); err != nil {
Michael Yang's avatar
Michael Yang committed
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
			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()
			}
		}

510
511
512
		return generateInteractive(cmd, opts)
	}
	return generate(cmd, opts)
Bruce MacDonald's avatar
Bruce MacDonald committed
513
514
}

515
func PushHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
516
	client, err := api.ClientFromEnvironment()
517
518
519
	if err != nil {
		return err
	}
520

521
522
523
524
525
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
526
527
528
529
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)
530
531
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
532

533
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
534
		if resp.Digest != "" {
535
536
537
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
538
539
540

			bar, ok := bars[resp.Digest]
			if !ok {
541
				bar = progress.NewBar(fmt.Sprintf("pushing %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
542
543
544
545
546
547
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
548
549
550
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
551
552
553
554
555
556

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

557
558
559
		return nil
	}

Michael Yang's avatar
Michael Yang committed
560
	request := api.PushRequest{Name: args[0], Insecure: insecure}
Michael Yang's avatar
Michael Yang committed
561
	if err := client.Push(cmd.Context(), &request, fn); err != nil {
562
563
564
565
566
567
		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
568
569
570
		return err
	}

571
	spinner.Stop()
Michael Yang's avatar
Michael Yang committed
572
	return nil
573
574
}

575
func ListHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
576
	client, err := api.ClientFromEnvironment()
577
578
579
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
580

Michael Yang's avatar
Michael Yang committed
581
	models, err := client.List(cmd.Context())
Patrick Devine's avatar
Patrick Devine committed
582
583
584
585
586
587
588
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
Michael Yang's avatar
Michael Yang committed
589
		if len(args) == 0 || strings.HasPrefix(m.Name, args[0]) {
590
			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
591
		}
Patrick Devine's avatar
Patrick Devine committed
592
593
594
	}

	table := tablewriter.NewWriter(os.Stdout)
Patrick Devine's avatar
Patrick Devine committed
595
	table.SetHeader([]string{"NAME", "ID", "SIZE", "MODIFIED"})
Patrick Devine's avatar
Patrick Devine committed
596
597
598
599
600
	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
601
	table.SetTablePadding("    ")
Patrick Devine's avatar
Patrick Devine committed
602
603
604
605
606
607
	table.AppendBulk(data)
	table.Render()

	return nil
}

608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
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
636
637
638
639
640
641
642
643
644

			var until string
			delta := time.Since(m.ExpiresAt)
			if delta > 0 {
				until = "Stopping..."
			} else {
				until = format.HumanTime(m.ExpiresAt, "Never")
			}
			data = append(data, []string{m.Name, m.Digest[:12], format.HumanBytes(m.Size), procStr, until})
645
646
647
648
649
650
651
652
653
654
		}
	}

	table := tablewriter.NewWriter(os.Stdout)
	table.SetHeader([]string{"NAME", "ID", "SIZE", "PROCESSOR", "UNTIL"})
	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
655
	table.SetTablePadding("    ")
656
657
658
659
660
661
	table.AppendBulk(data)
	table.Render()

	return nil
}

662
func DeleteHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
663
	client, err := api.ClientFromEnvironment()
664
665
666
	if err != nil {
		return err
	}
667

668
669
670
671
672
673
674
675
676
677
678
	// 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)
		}
	}

679
680
	for _, name := range args {
		req := api.DeleteRequest{Name: name}
Michael Yang's avatar
Michael Yang committed
681
		if err := client.Delete(cmd.Context(), &req); err != nil {
682
683
684
			return err
		}
		fmt.Printf("deleted '%s'\n", name)
685
686
687
688
	}
	return nil
}

Patrick Devine's avatar
Patrick Devine committed
689
func ShowHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
690
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
	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")

	for _, boolErr := range []error{errLicense, errModelfile, errParams, errSystem, errTemplate} {
		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 {
736
		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
737
738
	}

739
740
741
742
743
	req := api.ShowRequest{Name: args[0]}
	resp, err := client.Show(cmd.Context(), &req)
	if err != nil {
		return err
	}
744

745
	if flagsSet == 1 {
746
747
748
749
750
751
752
753
		switch showType {
		case "license":
			fmt.Println(resp.License)
		case "modelfile":
			fmt.Println(resp.Modelfile)
		case "parameters":
			fmt.Println(resp.Parameters)
		case "system":
754
			fmt.Print(resp.System)
755
		case "template":
756
			fmt.Print(resp.Template)
757
758
759
		}

		return nil
Patrick Devine's avatar
Patrick Devine committed
760
761
	}

Michael Yang's avatar
Michael Yang committed
762
	return showInfo(resp, os.Stdout)
763
764
}

Michael Yang's avatar
Michael Yang committed
765
766
767
768
769
770
771
772
func showInfo(resp *api.ShowResponse, w io.Writer) error {
	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("    ")
773

Michael Yang's avatar
Michael Yang committed
774
775
776
		switch header {
		case "Template", "System", "License":
			table.SetColWidth(100)
777
778
		}

Michael Yang's avatar
Michael Yang committed
779
780
781
		table.AppendBulk(rows())
		table.Render()
		fmt.Fprintln(w)
Patrick Devine's avatar
Patrick Devine committed
782
783
	}

Michael Yang's avatar
Michael Yang committed
784
785
786
787
788
789
790
791
792
793
794
795
796
797
	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
	})
798

Michael Yang's avatar
Michael Yang committed
799
800
801
802
803
804
805
806
807
	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
		})
808
809
	}

Michael Yang's avatar
Michael Yang committed
810
811
812
813
814
815
816
817
818
819
	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
		})
820
821
	}

Michael Yang's avatar
Michael Yang committed
822
823
824
825
826
	head := func(s string, n int) (rows [][]string) {
		scanner := bufio.NewScanner(strings.NewReader(s))
		for scanner.Scan() && (len(rows) < n || n < 0) {
			if text := scanner.Text(); text != "" {
				rows = append(rows, []string{"", strings.TrimSpace(text)})
827
828
			}
		}
Michael Yang's avatar
Michael Yang committed
829
		return
830
831
	}

Michael Yang's avatar
Michael Yang committed
832
833
834
835
836
	if resp.System != "" {
		tableRender("System", func() [][]string {
			return head(resp.System, 2)
		})
	}
837

Michael Yang's avatar
Michael Yang committed
838
839
840
841
	if resp.License != "" {
		tableRender("License", func() [][]string {
			return head(resp.License, 2)
		})
842
	}
Michael Yang's avatar
Michael Yang committed
843
844

	return nil
845
846
}

Patrick Devine's avatar
Patrick Devine committed
847
func CopyHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
848
	client, err := api.ClientFromEnvironment()
849
850
851
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
852
853

	req := api.CopyRequest{Source: args[0], Destination: args[1]}
Michael Yang's avatar
Michael Yang committed
854
	if err := client.Copy(cmd.Context(), &req); err != nil {
Patrick Devine's avatar
Patrick Devine committed
855
856
857
858
859
860
		return err
	}
	fmt.Printf("copied '%s' to '%s'\n", args[0], args[1])
	return nil
}

861
func PullHandler(cmd *cobra.Command, args []string) error {
862
863
864
865
866
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
867
	client, err := api.ClientFromEnvironment()
868
869
870
	if err != nil {
		return err
	}
871

Michael Yang's avatar
Michael Yang committed
872
873
874
875
876
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

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

877
878
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
879

880
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
881
		if resp.Digest != "" {
882
883
884
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
885
886
887

			bar, ok := bars[resp.Digest]
			if !ok {
888
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
889
890
891
892
893
894
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
895
896
897
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
898
899
900
901
902
903

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

904
905
		return nil
	}
906

Michael Yang's avatar
Michael Yang committed
907
	request := api.PullRequest{Name: args[0], Insecure: insecure}
Michael Yang's avatar
Michael Yang committed
908
	if err := client.Pull(cmd.Context(), &request, fn); err != nil {
Michael Yang's avatar
Michael Yang committed
909
910
911
912
		return err
	}

	return nil
Michael Yang's avatar
Michael Yang committed
913
914
}

915
916
type generateContextKey string

917
type runOptions struct {
918
919
920
921
922
923
924
925
926
927
	Model       string
	ParentModel string
	Prompt      string
	Messages    []api.Message
	WordWrap    bool
	Format      string
	System      string
	Images      []api.ImageData
	Options     map[string]interface{}
	MultiModal  bool
928
	KeepAlive   *api.Duration
929
930
}

931
932
933
934
935
936
937
938
939
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
940
941
			if state.lineLength+1 > termWidth-5 {
				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
942
943
944
945
946
947
948
					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
949
950
				a := runewidth.StringWidth(state.wordBuffer)
				if a > 0 {
951
					fmt.Printf("\x1b[%dD", a)
952
953
				}
				fmt.Printf("\x1b[K\n")
954
				fmt.Printf("%s%c", state.wordBuffer, ch)
955
956
957
				chWidth := runewidth.RuneWidth(ch)

				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
958
959
			} else {
				fmt.Print(string(ch))
960
961
962
963
				state.lineLength += runewidth.RuneWidth(ch)
				if runewidth.RuneWidth(ch) >= 2 {
					state.wordBuffer = ""
					continue
Josh Yan's avatar
Josh Yan committed
964
				}
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032

				switch ch {
				case ' ':
					state.wordBuffer = ""
				case '\n':
					state.lineLength = 0
				default:
					state.wordBuffer += string(ch)
				}
			}
		}
	} else {
		fmt.Printf("%s%s", state.wordBuffer, content)
		if len(state.wordBuffer) > 0 {
			state.wordBuffer = ""
		}
	}
}

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{}
	var latest api.ChatResponse
	var fullResponse strings.Builder
	var role string

	fn := func(response api.ChatResponse) error {
		p.StopAndClear()

		latest = response

		role = response.Message.Role
		content := response.Message.Content
		fullResponse.WriteString(content)

		displayResponse(content, opts.WordWrap, state)

		return nil
	}

	req := &api.ChatRequest{
		Model:    opts.Model,
		Messages: opts.Messages,
		Format:   opts.Format,
		Options:  opts.Options,
	}

1033
1034
1035
1036
	if opts.KeepAlive != nil {
		req.KeepAlive = opts.KeepAlive
	}

1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
	if err := client.Chat(cancelCtx, req, fn); err != nil {
		if errors.Is(err, context.Canceled) {
			return nil, nil
		}
		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
1062
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
1063
	if err != nil {
1064
		return err
Patrick Devine's avatar
Patrick Devine committed
1065
	}
Michael Yang's avatar
Michael Yang committed
1066

Michael Yang's avatar
Michael Yang committed
1067
	p := progress.NewProgress(os.Stderr)
1068
	defer p.StopAndClear()
1069

Michael Yang's avatar
Michael Yang committed
1070
1071
1072
	spinner := progress.NewSpinner("")
	p.Add("", spinner)

1073
1074
1075
1076
1077
1078
1079
	var latest api.GenerateResponse

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

Michael Yang's avatar
Michael Yang committed
1080
	ctx, cancel := context.WithCancel(cmd.Context())
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
	defer cancel()

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

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

1091
	var state *displayResponseState = &displayResponseState{}
1092

1093
	fn := func(response api.GenerateResponse) error {
Michael Yang's avatar
Michael Yang committed
1094
		p.StopAndClear()
1095

Patrick Devine's avatar
Patrick Devine committed
1096
		latest = response
1097
		content := response.Response
1098

1099
		displayResponse(content, opts.WordWrap, state)
1100

Patrick Devine's avatar
Patrick Devine committed
1101
1102
		return nil
	}
1103

1104
1105
1106
1107
1108
1109
1110
	if opts.MultiModal {
		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
		if err != nil {
			return err
		}
	}

Michael Yang's avatar
Michael Yang committed
1111
	request := api.GenerateRequest{
1112
1113
1114
1115
1116
1117
1118
1119
		Model:     opts.Model,
		Prompt:    opts.Prompt,
		Context:   generateContext,
		Images:    opts.Images,
		Format:    opts.Format,
		System:    opts.System,
		Options:   opts.Options,
		KeepAlive: opts.KeepAlive,
Michael Yang's avatar
Michael Yang committed
1120
1121
1122
	}

	if err := client.Generate(ctx, &request, fn); err != nil {
1123
		if errors.Is(err, context.Canceled) {
1124
			return nil
1125
		}
1126
		return err
Patrick Devine's avatar
Patrick Devine committed
1127
	}
1128

1129
	if opts.Prompt != "" {
Michael Yang's avatar
Michael Yang committed
1130
1131
		fmt.Println()
		fmt.Println()
Patrick Devine's avatar
Patrick Devine committed
1132
	}
1133

1134
1135
1136
1137
	if !latest.Done {
		return nil
	}

Patrick Devine's avatar
Patrick Devine committed
1138
1139
	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
1140
		return err
Patrick Devine's avatar
Patrick Devine committed
1141
	}
Michael Yang's avatar
Michael Yang committed
1142

Patrick Devine's avatar
Patrick Devine committed
1143
1144
	if verbose {
		latest.Summary()
Michael Yang's avatar
Michael Yang committed
1145
	}
Michael Yang's avatar
Michael Yang committed
1146

Patrick Devine's avatar
Patrick Devine committed
1147
1148
1149
	ctx = context.WithValue(cmd.Context(), generateContextKey("context"), latest.Context)
	cmd.SetContext(ctx)

1150
	return nil
Michael Yang's avatar
Michael Yang committed
1151
1152
}

1153
func RunServer(_ *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1154
	if err := initializeKeypair(); err != nil {
1155
1156
1157
		return err
	}

Michael Yang's avatar
host  
Michael Yang committed
1158
	ln, err := net.Listen("tcp", envconfig.Host().Host)
1159
1160
1161
	if err != nil {
		return err
	}
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1162

1163
1164
1165
1166
1167
1168
	err = server.Serve(ln)
	if errors.Is(err, http.ErrServerClosed) {
		return nil
	}

	return err
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1169
1170
}

1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
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
1183
		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
1184
1185
1186
1187
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1188
		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
1189
1190
1191
1192
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1193
		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
1194
1195
1196
			return fmt.Errorf("could not create directory %w", err)
		}

Michael Yang's avatar
Michael Yang committed
1197
		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
1198
1199
1200
			return err
		}

Michael Yang's avatar
Michael Yang committed
1201
		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
1202
1203
1204
1205
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1206
		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
1207

Michael Yang's avatar
Michael Yang committed
1208
		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
1209
1210
1211
			return err
		}

Michael Yang's avatar
Michael Yang committed
1212
		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
1213
1214
1215
1216
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1217
func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1218
	client, err := api.ClientFromEnvironment()
1219
1220
1221
	if err != nil {
		return err
	}
Michael Yang's avatar
Michael Yang committed
1222
	if err := client.Heartbeat(cmd.Context()); err != nil {
1223
		if !strings.Contains(err.Error(), " refused") {
Bruce MacDonald's avatar
Bruce MacDonald committed
1224
1225
			return err
		}
1226
		if err := startApp(cmd.Context(), client); err != nil {
Michael Yang's avatar
lint  
Michael Yang committed
1227
			return errors.New("could not connect to ollama app, is it running?")
1228
1229
1230
1231
1232
		}
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1233
1234
1235
1236
1237
1238
1239
1240
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
1241
1242
1243
1244
1245
		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
1246
1247
	}

1248
	if serverVersion != version.Version {
Michael Yang's avatar
Michael Yang committed
1249
		fmt.Printf("Warning: client version is %s\n", version.Version)
1250
	}
Michael Yang's avatar
Michael Yang committed
1251
1252
}

1253
func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
1254
1255
1256
1257
1258
	if len(envs) == 0 {
		return
	}

	envUsage := `
1259
1260
Environment Variables:
`
1261
	for _, e := range envs {
1262
		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
1263
1264
1265
	}

	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
1266
1267
}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1268
1269
func NewCLI() *cobra.Command {
	log.SetFlags(log.LstdFlags | log.Lshortfile)
Michael Yang's avatar
Michael Yang committed
1270
	cobra.EnableCommandSorting = false
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1271

1272
	if runtime.GOOS == "windows" && term.IsTerminal(int(os.Stdout.Fd())) {
1273
		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
1274
1275
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1276
	rootCmd := &cobra.Command{
1277
1278
1279
1280
		Use:           "ollama",
		Short:         "Large language model runner",
		SilenceUsage:  true,
		SilenceErrors: true,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1281
1282
1283
		CompletionOptions: cobra.CompletionOptions{
			DisableDefaultCmd: true,
		},
Michael Yang's avatar
Michael Yang committed
1284
1285
1286
1287
1288
1289
1290
1291
		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
1292
1293
	}

Michael Yang's avatar
Michael Yang committed
1294
	rootCmd.Flags().BoolP("version", "v", false, "Show version information")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1295

1296
	createCmd := &cobra.Command{
1297
1298
		Use:     "create MODEL",
		Short:   "Create a model from a Modelfile",
Michael Yang's avatar
Michael Yang committed
1299
		Args:    cobra.ExactArgs(1),
1300
1301
		PreRunE: checkServerHeartbeat,
		RunE:    CreateHandler,
1302
1303
	}

1304
	createCmd.Flags().StringP("file", "f", "", "Name of the Modelfile (default \"Modelfile\"")
1305
	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_0)")
1306

Patrick Devine's avatar
Patrick Devine committed
1307
1308
1309
	showCmd := &cobra.Command{
		Use:     "show MODEL",
		Short:   "Show information for a model",
Michael Yang's avatar
Michael Yang committed
1310
		Args:    cobra.ExactArgs(1),
Patrick Devine's avatar
Patrick Devine committed
1311
1312
1313
1314
1315
1316
1317
1318
		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")
1319
	showCmd.Flags().Bool("system", false, "Show system message of a model")
Patrick Devine's avatar
Patrick Devine committed
1320

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1321
	runCmd := &cobra.Command{
1322
1323
1324
1325
1326
		Use:     "run MODEL [PROMPT]",
		Short:   "Run a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    RunHandler,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1327
1328
	}

1329
	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
1330
	runCmd.Flags().Bool("verbose", false, "Show timings for response")
1331
	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
1332
	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1333
	runCmd.Flags().String("format", "", "Response format (e.g. json)")
Patrick Devine's avatar
Patrick Devine committed
1334
1335
1336
1337
1338
1339
1340
1341
1342

	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
1343
1344
1345
1346
	serveCmd := &cobra.Command{
		Use:     "serve",
		Aliases: []string{"start"},
		Short:   "Start ollama",
Michael Yang's avatar
Michael Yang committed
1347
		Args:    cobra.ExactArgs(0),
Michael Yang's avatar
Michael Yang committed
1348
		RunE:    RunServer,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1349
1350
	}

1351
	pullCmd := &cobra.Command{
1352
1353
		Use:     "pull MODEL",
		Short:   "Pull a model from a registry",
Michael Yang's avatar
Michael Yang committed
1354
		Args:    cobra.ExactArgs(1),
1355
1356
		PreRunE: checkServerHeartbeat,
		RunE:    PullHandler,
1357
1358
	}

1359
1360
	pullCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1361
	pushCmd := &cobra.Command{
1362
1363
		Use:     "push MODEL",
		Short:   "Push a model to a registry",
Michael Yang's avatar
Michael Yang committed
1364
		Args:    cobra.ExactArgs(1),
1365
1366
		PreRunE: checkServerHeartbeat,
		RunE:    PushHandler,
1367
1368
	}

1369
1370
	pushCmd.Flags().Bool("insecure", false, "Use an insecure registry")

Patrick Devine's avatar
Patrick Devine committed
1371
	listCmd := &cobra.Command{
1372
		Use:     "list",
Patrick Devine's avatar
Patrick Devine committed
1373
		Aliases: []string{"ls"},
1374
		Short:   "List models",
1375
		PreRunE: checkServerHeartbeat,
1376
		RunE:    ListHandler,
1377
	}
1378
1379
1380
1381
1382
1383
1384
1385

	psCmd := &cobra.Command{
		Use:     "ps",
		Short:   "List running models",
		PreRunE: checkServerHeartbeat,
		RunE:    ListRunningHandler,
	}

Patrick Devine's avatar
Patrick Devine committed
1386
	copyCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1387
		Use:     "cp SOURCE DESTINATION",
1388
		Short:   "Copy a model",
Michael Yang's avatar
Michael Yang committed
1389
		Args:    cobra.ExactArgs(2),
1390
1391
		PreRunE: checkServerHeartbeat,
		RunE:    CopyHandler,
Patrick Devine's avatar
Patrick Devine committed
1392
1393
	}

1394
	deleteCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1395
		Use:     "rm MODEL [MODEL...]",
1396
1397
1398
1399
		Short:   "Remove a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    DeleteHandler,
Patrick Devine's avatar
Patrick Devine committed
1400
1401
	}

1402
1403
1404
	envVars := envconfig.AsMap()

	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
1405

1406
1407
1408
1409
	for _, cmd := range []*cobra.Command{
		createCmd,
		showCmd,
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1410
		stopCmd,
1411
1412
1413
		pullCmd,
		pushCmd,
		listCmd,
1414
		psCmd,
1415
1416
		copyCmd,
		deleteCmd,
1417
		serveCmd,
1418
	} {
1419
1420
		switch cmd {
		case runCmd:
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
			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"],
1433
				envVars["OLLAMA_SCHED_SPREAD"],
1434
				envVars["OLLAMA_TMPDIR"],
1435
1436
				envVars["OLLAMA_FLASH_ATTENTION"],
				envVars["OLLAMA_LLM_LIBRARY"],
1437
				envVars["OLLAMA_GPU_OVERHEAD"],
1438
				envVars["OLLAMA_LOAD_TIMEOUT"],
1439
			})
1440
1441
1442
		default:
			appendEnvDocs(cmd, envs)
		}
1443
1444
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1445
1446
	rootCmd.AddCommand(
		serveCmd,
1447
		createCmd,
Patrick Devine's avatar
Patrick Devine committed
1448
		showCmd,
1449
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1450
		stopCmd,
1451
1452
		pullCmd,
		pushCmd,
Patrick Devine's avatar
Patrick Devine committed
1453
		listCmd,
1454
		psCmd,
Patrick Devine's avatar
Patrick Devine committed
1455
		copyCmd,
1456
		deleteCmd,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1457
1458
1459
1460
	)

	return rootCmd
}