cmd.go 34.1 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/json"
12
	"encoding/pem"
Michael Yang's avatar
Michael Yang committed
13
	"errors"
Bruce MacDonald's avatar
Bruce MacDonald committed
14
	"fmt"
Michael Yang's avatar
Michael Yang committed
15
	"io"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
16
	"log"
17
	"math"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
18
	"net"
19
	"net/http"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
20
	"os"
21
	"os/signal"
22
	"path/filepath"
23
	"runtime"
Michael Yang's avatar
Michael Yang committed
24
	"strconv"
Michael Yang's avatar
Michael Yang committed
25
	"strings"
26
	"sync/atomic"
27
	"syscall"
Michael Yang's avatar
Michael Yang committed
28
	"time"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
29

30
	"github.com/containerd/console"
31
	"github.com/mattn/go-runewidth"
Patrick Devine's avatar
Patrick Devine committed
32
	"github.com/olekukonko/tablewriter"
Michael Yang's avatar
Michael Yang committed
33
	"github.com/spf13/cobra"
34
	"golang.org/x/crypto/ssh"
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
42
	"github.com/ollama/ollama/progress"
	"github.com/ollama/ollama/server"
43
	"github.com/ollama/ollama/types/model"
44
	"github.com/ollama/ollama/version"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
45
46
)

47
48
49
50
51
52
53
54
55
56
57
58
59
60
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)
61
	if err != nil {
62
		return "", err
63
64
	}

65
	_, err = os.Stat(absName)
66
	if err != nil {
67
		return fn, err
68
	}
69

70
71
72
73
	return absName, nil
}

func CreateHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
74
75
76
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

77
78
79
80
81
82
83
84
85
86
	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
87
		return err
88
89
90
91
92
93
94
95
	} else {
		f, err := os.Open(filename)
		if err != nil {
			return err
		}

		reader = f
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
96
97
	}

98
	modelfile, err := parser.ParseFile(reader)
Michael Yang's avatar
Michael Yang committed
99
100
101
102
103
104
105
106
107
	if err != nil {
		return err
	}

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

108
109
	status := "transferring model data"
	spinner := progress.NewSpinner(status)
110
	p.Add(status, spinner)
111
	defer p.Stop()
112

113
114
115
116
117
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
118
119
	for i := range modelfile.Commands {
		switch modelfile.Commands[i].Name {
Michael Yang's avatar
Michael Yang committed
120
		case "model", "adapter":
Michael Yang's avatar
Michael Yang committed
121
			path := modelfile.Commands[i].Args
Michael Yang's avatar
Michael Yang committed
122
123
124
125
126
127
			if path == "~" {
				path = home
			} else if strings.HasPrefix(path, "~/") {
				path = filepath.Join(home, path[2:])
			}

128
129
130
131
			if !filepath.IsAbs(path) {
				path = filepath.Join(filepath.Dir(filename), path)
			}

132
			fi, err := os.Stat(path)
Michael Yang's avatar
Michael Yang committed
133
			if errors.Is(err, os.ErrNotExist) && modelfile.Commands[i].Name == "model" {
Michael Yang's avatar
Michael Yang committed
134
				continue
Michael Yang's avatar
Michael Yang committed
135
136
137
138
			} else if err != nil {
				return err
			}

139
			if fi.IsDir() {
Michael Yang's avatar
Michael Yang committed
140
141
142
				// this is likely a safetensors or pytorch directory
				// TODO make this work w/ adapters
				tempfile, err := tempZipFiles(path)
143
144
145
				if err != nil {
					return err
				}
Michael Yang's avatar
Michael Yang committed
146
				defer os.RemoveAll(tempfile)
147

Michael Yang's avatar
Michael Yang committed
148
				path = tempfile
Michael Yang's avatar
Michael Yang committed
149
150
			}

151
			digest, err := createBlob(cmd, client, path, spinner)
152
			if err != nil {
Michael Yang's avatar
Michael Yang committed
153
154
155
				return err
			}

Michael Yang's avatar
Michael Yang committed
156
			modelfile.Commands[i].Args = "@" + digest
Michael Yang's avatar
Michael Yang committed
157
158
		}
	}
Michael Yang's avatar
Michael Yang committed
159

Michael Yang's avatar
Michael Yang committed
160
	bars := make(map[string]*progress.Bar)
161
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
162
163
164
165
166
		if resp.Digest != "" {
			spinner.Stop()

			bar, ok := bars[resp.Digest]
			if !ok {
167
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
168
169
170
171
172
173
174
175
176
177
178
179
180
				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)
		}

181
182
183
		return nil
	}

184
	quantize, _ := cmd.Flags().GetString("quantize")
Michael Yang's avatar
Michael Yang committed
185

186
	request := api.CreateRequest{Name: args[0], Modelfile: modelfile.String(), Quantize: quantize}
Michael Yang's avatar
Michael Yang committed
187
	if err := client.Create(cmd.Context(), &request, fn); err != nil {
188
189
190
191
192
193
		return err
	}

	return nil
}

Michael Yang's avatar
Michael Yang committed
194
195
196
197
198
199
200
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
201
202
	detectContentType := func(path string) (string, error) {
		f, err := os.Open(path)
Michael Yang's avatar
Michael Yang committed
203
204
205
		if err != nil {
			return "", err
		}
Michael Yang's avatar
Michael Yang committed
206
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
207

Michael Yang's avatar
Michael Yang committed
208
209
		var b bytes.Buffer
		b.Grow(512)
Michael Yang's avatar
Michael Yang committed
210

Michael Yang's avatar
Michael Yang committed
211
212
213
214
215
216
		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
217
218
	}

Michael Yang's avatar
Michael Yang committed
219
220
221
222
223
224
225
226
227
228
229
	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
230
			}
Michael Yang's avatar
Michael Yang committed
231
232
233
234
235
236
237
238
239
240
		}

		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...)
241
242
243
244
245
246
	} 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
247
248
249
250
	} 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
251
	} else if pt, _ := glob(filepath.Join(path, "consolidated*.pth"), "application/zip"); len(pt) > 0 {
Michael Yang's avatar
Michael Yang committed
252
253
254
255
		// pytorch files might also be unresolved git lfs references; skip if they are
		// covers consolidated.x.pth, consolidated.pth
		files = append(files, pt...)
	} else {
256
		return "", errModelNotFound
Michael Yang's avatar
Michael Yang committed
257
258
259
260
261
262
263
264
265
	}

	// 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
266
267
268
269
270
271
272
273
	// 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
274
275
276
277
278
279
280
281
282
	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...)
	}

283
284
285
	zipfile := zip.NewWriter(tempfile)
	defer zipfile.Close()

Michael Yang's avatar
Michael Yang committed
286
287
288
	for _, file := range files {
		f, err := os.Open(file)
		if err != nil {
Michael Yang's avatar
Michael Yang committed
289
290
			return "", err
		}
Michael Yang's avatar
Michael Yang committed
291
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
292
293
294
295
296
297

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

Michael Yang's avatar
Michael Yang committed
298
		zfi, err := zip.FileInfoHeader(fi)
Michael Yang's avatar
Michael Yang committed
299
300
301
302
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
303
304
305
306
307
		zfi.Name, err = filepath.Rel(path, file)
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
308
		zf, err := zipfile.CreateHeader(zfi)
Michael Yang's avatar
Michael Yang committed
309
310
311
312
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
313
		if _, err := io.Copy(zf, f); err != nil {
Michael Yang's avatar
Michael Yang committed
314
315
316
317
318
319
320
			return "", err
		}
	}

	return tempfile.Name(), nil
}

321
func createBlob(cmd *cobra.Command, client *api.Client, path string, spinner *progress.Spinner) (string, error) {
322
323
324
325
326
327
	bin, err := os.Open(path)
	if err != nil {
		return "", err
	}
	defer bin.Close()

328
329
330
331
332
333
334
	// Get file info to retrieve the size
	fileInfo, err := bin.Stat()
	if err != nil {
		return "", err
	}
	fileSize := fileInfo.Size()

335
336
337
338
	hash := sha256.New()
	if _, err := io.Copy(hash, bin); err != nil {
		return "", err
	}
339
340
341
342

	if _, err := bin.Seek(0, io.SeekStart); err != nil {
		return "", err
	}
343

344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
	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
			}
		}
	}()

365
	digest := fmt.Sprintf("sha256:%x", hash.Sum(nil))
366
	if err = client.CreateBlob(cmd.Context(), digest, io.TeeReader(bin, &pw)); err != nil {
367
368
369
370
371
		return "", err
	}
	return digest, nil
}

372
373
374
375
376
377
378
379
380
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
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
412
413
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
}

414
func RunHandler(cmd *cobra.Command, args []string) error {
415
416
417
	interactive := true

	opts := runOptions{
418
419
420
		Model:    args[0],
		WordWrap: os.Getenv("TERM") == "xterm-256color",
		Options:  map[string]interface{}{},
421
422
423
424
425
426
427
428
	}

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

429
430
431
432
433
434
435
436
437
438
439
440
	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}
	}

441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
	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
	}
457
458
459
460
	// Be quiet if we're redirecting to a pipe or file
	if !term.IsTerminal(int(os.Stdout.Fd())) {
		interactive = false
	}
461
462
463
464
465
466
467

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

468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
	// 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
490
491
	}

492
	opts.MultiModal = len(info.ProjectorInfo) != 0
493
494
495
	opts.ParentModel = info.Details.ParentModel

	if interactive {
Patrick Devine's avatar
Patrick Devine committed
496
		if err := loadOrUnloadModel(cmd, &opts); err != nil {
Michael Yang's avatar
Michael Yang committed
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
			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()
			}
		}

512
513
514
		return generateInteractive(cmd, opts)
	}
	return generate(cmd, opts)
Bruce MacDonald's avatar
Bruce MacDonald committed
515
516
}

517
func PushHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
518
	client, err := api.ClientFromEnvironment()
519
520
521
	if err != nil {
		return err
	}
522

523
524
525
526
527
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
528
529
530
531
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)
532
533
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
534

535
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
536
		if resp.Digest != "" {
537
538
539
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
540
541
542

			bar, ok := bars[resp.Digest]
			if !ok {
543
				bar = progress.NewBar(fmt.Sprintf("pushing %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
544
545
546
547
548
549
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
550
551
552
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
553
554
555
556
557
558

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

559
560
561
		return nil
	}

Michael Yang's avatar
Michael Yang committed
562
	request := api.PushRequest{Name: args[0], Insecure: insecure}
563
564

	n := model.ParseName(args[0])
Michael Yang's avatar
Michael Yang committed
565
	if err := client.Push(cmd.Context(), &request, fn); err != nil {
566
567
568
569
570
571
		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
572
573
574
		return err
	}

575
	p.Stop()
576
	spinner.Stop()
577
578
579
580
581
582
583
584

	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
585
	return nil
586
587
}

588
func ListHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
589
	client, err := api.ClientFromEnvironment()
590
591
592
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
593

Michael Yang's avatar
Michael Yang committed
594
	models, err := client.List(cmd.Context())
Patrick Devine's avatar
Patrick Devine committed
595
596
597
598
599
600
601
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
Michael Yang's avatar
Michael Yang committed
602
		if len(args) == 0 || strings.HasPrefix(m.Name, args[0]) {
603
			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
604
		}
Patrick Devine's avatar
Patrick Devine committed
605
606
607
	}

	table := tablewriter.NewWriter(os.Stdout)
Patrick Devine's avatar
Patrick Devine committed
608
	table.SetHeader([]string{"NAME", "ID", "SIZE", "MODIFIED"})
Patrick Devine's avatar
Patrick Devine committed
609
610
611
612
613
	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
614
	table.SetTablePadding("    ")
Patrick Devine's avatar
Patrick Devine committed
615
616
617
618
619
620
	table.AppendBulk(data)
	table.Render()

	return nil
}

621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
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
649
650
651
652
653
654
655
656
657

			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})
658
659
660
661
662
663
664
665
666
667
		}
	}

	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
668
	table.SetTablePadding("    ")
669
670
671
672
673
674
	table.AppendBulk(data)
	table.Render()

	return nil
}

675
func DeleteHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
676
	client, err := api.ClientFromEnvironment()
677
678
679
	if err != nil {
		return err
	}
680

681
682
683
684
685
686
687
688
689
690
691
	// 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)
		}
	}

692
693
	for _, name := range args {
		req := api.DeleteRequest{Name: name}
Michael Yang's avatar
Michael Yang committed
694
		if err := client.Delete(cmd.Context(), &req); err != nil {
695
696
697
			return err
		}
		fmt.Printf("deleted '%s'\n", name)
698
699
700
701
	}
	return nil
}

Patrick Devine's avatar
Patrick Devine committed
702
func ShowHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
703
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
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
736
737
738
739
740
741
742
743
744
745
746
747
748
	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 {
749
		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
750
751
	}

752
753
754
755
756
	req := api.ShowRequest{Name: args[0]}
	resp, err := client.Show(cmd.Context(), &req)
	if err != nil {
		return err
	}
757

758
	if flagsSet == 1 {
759
760
761
762
763
764
765
766
		switch showType {
		case "license":
			fmt.Println(resp.License)
		case "modelfile":
			fmt.Println(resp.Modelfile)
		case "parameters":
			fmt.Println(resp.Parameters)
		case "system":
767
			fmt.Print(resp.System)
768
		case "template":
769
			fmt.Print(resp.Template)
770
771
772
		}

		return nil
Patrick Devine's avatar
Patrick Devine committed
773
774
	}

Michael Yang's avatar
Michael Yang committed
775
	return showInfo(resp, os.Stdout)
776
777
}

Michael Yang's avatar
Michael Yang committed
778
779
780
781
782
783
784
785
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("    ")
786

Michael Yang's avatar
Michael Yang committed
787
788
789
		switch header {
		case "Template", "System", "License":
			table.SetColWidth(100)
790
791
		}

Michael Yang's avatar
Michael Yang committed
792
793
794
		table.AppendBulk(rows())
		table.Render()
		fmt.Fprintln(w)
Patrick Devine's avatar
Patrick Devine committed
795
796
	}

Michael Yang's avatar
Michael Yang committed
797
798
799
800
801
802
803
804
805
806
807
808
809
810
	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
	})
811

Michael Yang's avatar
Michael Yang committed
812
813
814
815
816
817
818
819
820
	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
		})
821
822
	}

Michael Yang's avatar
Michael Yang committed
823
824
825
826
827
828
829
830
831
832
	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
		})
833
834
	}

Michael Yang's avatar
Michael Yang committed
835
836
837
838
839
	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)})
840
841
			}
		}
Michael Yang's avatar
Michael Yang committed
842
		return
843
844
	}

Michael Yang's avatar
Michael Yang committed
845
846
847
848
849
	if resp.System != "" {
		tableRender("System", func() [][]string {
			return head(resp.System, 2)
		})
	}
850

Michael Yang's avatar
Michael Yang committed
851
852
853
854
	if resp.License != "" {
		tableRender("License", func() [][]string {
			return head(resp.License, 2)
		})
855
	}
Michael Yang's avatar
Michael Yang committed
856
857

	return nil
858
859
}

Patrick Devine's avatar
Patrick Devine committed
860
func CopyHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
861
	client, err := api.ClientFromEnvironment()
862
863
864
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
865
866

	req := api.CopyRequest{Source: args[0], Destination: args[1]}
Michael Yang's avatar
Michael Yang committed
867
	if err := client.Copy(cmd.Context(), &req); err != nil {
Patrick Devine's avatar
Patrick Devine committed
868
869
870
871
872
873
		return err
	}
	fmt.Printf("copied '%s' to '%s'\n", args[0], args[1])
	return nil
}

874
func PullHandler(cmd *cobra.Command, args []string) error {
875
876
877
878
879
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
880
	client, err := api.ClientFromEnvironment()
881
882
883
	if err != nil {
		return err
	}
884

Michael Yang's avatar
Michael Yang committed
885
886
887
888
889
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

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

890
891
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
892

893
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
894
		if resp.Digest != "" {
895
896
897
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
898
899
900

			bar, ok := bars[resp.Digest]
			if !ok {
901
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
902
903
904
905
906
907
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
908
909
910
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
911
912
913
914
915
916

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

917
918
		return nil
	}
919

Michael Yang's avatar
Michael Yang committed
920
	request := api.PullRequest{Name: args[0], Insecure: insecure}
Michael Yang's avatar
Michael Yang committed
921
	if err := client.Pull(cmd.Context(), &request, fn); err != nil {
Michael Yang's avatar
Michael Yang committed
922
923
924
925
		return err
	}

	return nil
Michael Yang's avatar
Michael Yang committed
926
927
}

928
929
type generateContextKey string

930
type runOptions struct {
931
932
933
934
935
936
937
938
939
940
	Model       string
	ParentModel string
	Prompt      string
	Messages    []api.Message
	WordWrap    bool
	Format      string
	System      string
	Images      []api.ImageData
	Options     map[string]interface{}
	MultiModal  bool
941
	KeepAlive   *api.Duration
942
943
}

944
945
946
947
948
949
950
951
952
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
953
954
			if state.lineLength+1 > termWidth-5 {
				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
955
956
957
958
959
960
961
					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
962
963
				a := runewidth.StringWidth(state.wordBuffer)
				if a > 0 {
964
					fmt.Printf("\x1b[%dD", a)
965
966
				}
				fmt.Printf("\x1b[K\n")
967
				fmt.Printf("%s%c", state.wordBuffer, ch)
968
969
970
				chWidth := runewidth.RuneWidth(ch)

				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
971
972
			} else {
				fmt.Print(string(ch))
973
974
975
976
				state.lineLength += runewidth.RuneWidth(ch)
				if runewidth.RuneWidth(ch) >= 2 {
					state.wordBuffer = ""
					continue
Josh Yan's avatar
Josh Yan committed
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
1033
1034
1035
1036
1037
1038
1039
1040
1041

				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,
1042
		Format:   json.RawMessage(opts.Format),
1043
1044
1045
		Options:  opts.Options,
	}

1046
1047
1048
1049
	if opts.KeepAlive != nil {
		req.KeepAlive = opts.KeepAlive
	}

1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
	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
1075
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
1076
	if err != nil {
1077
		return err
Patrick Devine's avatar
Patrick Devine committed
1078
	}
Michael Yang's avatar
Michael Yang committed
1079

Michael Yang's avatar
Michael Yang committed
1080
	p := progress.NewProgress(os.Stderr)
1081
	defer p.StopAndClear()
1082

Michael Yang's avatar
Michael Yang committed
1083
1084
1085
	spinner := progress.NewSpinner("")
	p.Add("", spinner)

1086
1087
1088
1089
1090
1091
1092
	var latest api.GenerateResponse

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

Michael Yang's avatar
Michael Yang committed
1093
	ctx, cancel := context.WithCancel(cmd.Context())
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
	defer cancel()

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

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

1104
	var state *displayResponseState = &displayResponseState{}
1105

1106
	fn := func(response api.GenerateResponse) error {
Michael Yang's avatar
Michael Yang committed
1107
		p.StopAndClear()
1108

Patrick Devine's avatar
Patrick Devine committed
1109
		latest = response
1110
		content := response.Response
1111

1112
		displayResponse(content, opts.WordWrap, state)
1113

Patrick Devine's avatar
Patrick Devine committed
1114
1115
		return nil
	}
1116

1117
1118
1119
1120
1121
1122
1123
	if opts.MultiModal {
		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
		if err != nil {
			return err
		}
	}

Michael Yang's avatar
Michael Yang committed
1124
	request := api.GenerateRequest{
1125
1126
1127
1128
1129
1130
1131
1132
		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
1133
1134
1135
	}

	if err := client.Generate(ctx, &request, fn); err != nil {
1136
		if errors.Is(err, context.Canceled) {
1137
			return nil
1138
		}
1139
		return err
Patrick Devine's avatar
Patrick Devine committed
1140
	}
1141

1142
	if opts.Prompt != "" {
Michael Yang's avatar
Michael Yang committed
1143
1144
		fmt.Println()
		fmt.Println()
Patrick Devine's avatar
Patrick Devine committed
1145
	}
1146

1147
1148
1149
1150
	if !latest.Done {
		return nil
	}

Patrick Devine's avatar
Patrick Devine committed
1151
1152
	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
1153
		return err
Patrick Devine's avatar
Patrick Devine committed
1154
	}
Michael Yang's avatar
Michael Yang committed
1155

Patrick Devine's avatar
Patrick Devine committed
1156
1157
	if verbose {
		latest.Summary()
Michael Yang's avatar
Michael Yang committed
1158
	}
Michael Yang's avatar
Michael Yang committed
1159

Patrick Devine's avatar
Patrick Devine committed
1160
1161
1162
	ctx = context.WithValue(cmd.Context(), generateContextKey("context"), latest.Context)
	cmd.SetContext(ctx)

1163
	return nil
Michael Yang's avatar
Michael Yang committed
1164
1165
}

1166
func RunServer(_ *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1167
	if err := initializeKeypair(); err != nil {
1168
1169
1170
		return err
	}

Michael Yang's avatar
host  
Michael Yang committed
1171
	ln, err := net.Listen("tcp", envconfig.Host().Host)
1172
1173
1174
	if err != nil {
		return err
	}
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1175

1176
1177
1178
1179
1180
1181
	err = server.Serve(ln)
	if errors.Is(err, http.ErrServerClosed) {
		return nil
	}

	return err
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1182
1183
}

1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
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
1196
		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
1197
1198
1199
1200
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1201
		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
1202
1203
1204
1205
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1206
		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
1207
1208
1209
			return fmt.Errorf("could not create directory %w", err)
		}

Michael Yang's avatar
Michael Yang committed
1210
		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
1211
1212
1213
			return err
		}

Michael Yang's avatar
Michael Yang committed
1214
		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
1215
1216
1217
1218
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1219
		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
1220

Michael Yang's avatar
Michael Yang committed
1221
		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
1222
1223
1224
			return err
		}

Michael Yang's avatar
Michael Yang committed
1225
		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
1226
1227
1228
1229
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1230
func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1231
	client, err := api.ClientFromEnvironment()
1232
1233
1234
	if err != nil {
		return err
	}
Michael Yang's avatar
Michael Yang committed
1235
	if err := client.Heartbeat(cmd.Context()); err != nil {
1236
		if !strings.Contains(err.Error(), " refused") {
Bruce MacDonald's avatar
Bruce MacDonald committed
1237
1238
			return err
		}
1239
		if err := startApp(cmd.Context(), client); err != nil {
Michael Yang's avatar
lint  
Michael Yang committed
1240
			return errors.New("could not connect to ollama app, is it running?")
1241
1242
1243
1244
1245
		}
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1246
1247
1248
1249
1250
1251
1252
1253
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
1254
1255
1256
1257
1258
		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
1259
1260
	}

1261
	if serverVersion != version.Version {
Michael Yang's avatar
Michael Yang committed
1262
		fmt.Printf("Warning: client version is %s\n", version.Version)
1263
	}
Michael Yang's avatar
Michael Yang committed
1264
1265
}

1266
func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
1267
1268
1269
1270
1271
	if len(envs) == 0 {
		return
	}

	envUsage := `
1272
1273
Environment Variables:
`
1274
	for _, e := range envs {
1275
		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
1276
1277
1278
	}

	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
1279
1280
}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1281
1282
func NewCLI() *cobra.Command {
	log.SetFlags(log.LstdFlags | log.Lshortfile)
Michael Yang's avatar
Michael Yang committed
1283
	cobra.EnableCommandSorting = false
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1284

1285
	if runtime.GOOS == "windows" && term.IsTerminal(int(os.Stdout.Fd())) {
1286
		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
1287
1288
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1289
	rootCmd := &cobra.Command{
1290
1291
1292
1293
		Use:           "ollama",
		Short:         "Large language model runner",
		SilenceUsage:  true,
		SilenceErrors: true,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1294
1295
1296
		CompletionOptions: cobra.CompletionOptions{
			DisableDefaultCmd: true,
		},
Michael Yang's avatar
Michael Yang committed
1297
1298
1299
1300
1301
1302
1303
1304
		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
1305
1306
	}

Michael Yang's avatar
Michael Yang committed
1307
	rootCmd.Flags().BoolP("version", "v", false, "Show version information")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1308

1309
	createCmd := &cobra.Command{
1310
1311
		Use:     "create MODEL",
		Short:   "Create a model from a Modelfile",
Michael Yang's avatar
Michael Yang committed
1312
		Args:    cobra.ExactArgs(1),
1313
1314
		PreRunE: checkServerHeartbeat,
		RunE:    CreateHandler,
1315
1316
	}

1317
	createCmd.Flags().StringP("file", "f", "", "Name of the Modelfile (default \"Modelfile\"")
1318
	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_0)")
1319

Patrick Devine's avatar
Patrick Devine committed
1320
1321
1322
	showCmd := &cobra.Command{
		Use:     "show MODEL",
		Short:   "Show information for a model",
Michael Yang's avatar
Michael Yang committed
1323
		Args:    cobra.ExactArgs(1),
Patrick Devine's avatar
Patrick Devine committed
1324
1325
1326
1327
1328
1329
1330
1331
		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")
1332
	showCmd.Flags().Bool("system", false, "Show system message of a model")
Patrick Devine's avatar
Patrick Devine committed
1333

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1334
	runCmd := &cobra.Command{
1335
1336
1337
1338
1339
		Use:     "run MODEL [PROMPT]",
		Short:   "Run a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    RunHandler,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1340
1341
	}

1342
	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
1343
	runCmd.Flags().Bool("verbose", false, "Show timings for response")
1344
	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
1345
	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1346
	runCmd.Flags().String("format", "", "Response format (e.g. json)")
Patrick Devine's avatar
Patrick Devine committed
1347
1348
1349
1350
1351
1352
1353
1354
1355

	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
1356
1357
1358
1359
	serveCmd := &cobra.Command{
		Use:     "serve",
		Aliases: []string{"start"},
		Short:   "Start ollama",
Michael Yang's avatar
Michael Yang committed
1360
		Args:    cobra.ExactArgs(0),
Michael Yang's avatar
Michael Yang committed
1361
		RunE:    RunServer,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1362
1363
	}

1364
	pullCmd := &cobra.Command{
1365
1366
		Use:     "pull MODEL",
		Short:   "Pull a model from a registry",
Michael Yang's avatar
Michael Yang committed
1367
		Args:    cobra.ExactArgs(1),
1368
1369
		PreRunE: checkServerHeartbeat,
		RunE:    PullHandler,
1370
1371
	}

1372
1373
	pullCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1374
	pushCmd := &cobra.Command{
1375
1376
		Use:     "push MODEL",
		Short:   "Push a model to a registry",
Michael Yang's avatar
Michael Yang committed
1377
		Args:    cobra.ExactArgs(1),
1378
1379
		PreRunE: checkServerHeartbeat,
		RunE:    PushHandler,
1380
1381
	}

1382
1383
	pushCmd.Flags().Bool("insecure", false, "Use an insecure registry")

Patrick Devine's avatar
Patrick Devine committed
1384
	listCmd := &cobra.Command{
1385
		Use:     "list",
Patrick Devine's avatar
Patrick Devine committed
1386
		Aliases: []string{"ls"},
1387
		Short:   "List models",
1388
		PreRunE: checkServerHeartbeat,
1389
		RunE:    ListHandler,
1390
	}
1391
1392
1393
1394
1395
1396
1397
1398

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

Patrick Devine's avatar
Patrick Devine committed
1399
	copyCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1400
		Use:     "cp SOURCE DESTINATION",
1401
		Short:   "Copy a model",
Michael Yang's avatar
Michael Yang committed
1402
		Args:    cobra.ExactArgs(2),
1403
1404
		PreRunE: checkServerHeartbeat,
		RunE:    CopyHandler,
Patrick Devine's avatar
Patrick Devine committed
1405
1406
	}

1407
	deleteCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1408
		Use:     "rm MODEL [MODEL...]",
1409
1410
1411
1412
		Short:   "Remove a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    DeleteHandler,
Patrick Devine's avatar
Patrick Devine committed
1413
1414
	}

1415
1416
1417
	envVars := envconfig.AsMap()

	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
1418

1419
1420
1421
1422
	for _, cmd := range []*cobra.Command{
		createCmd,
		showCmd,
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1423
		stopCmd,
1424
1425
1426
		pullCmd,
		pushCmd,
		listCmd,
1427
		psCmd,
1428
1429
		copyCmd,
		deleteCmd,
1430
		serveCmd,
1431
	} {
1432
1433
		switch cmd {
		case runCmd:
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
			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"],
1446
				envVars["OLLAMA_SCHED_SPREAD"],
1447
				envVars["OLLAMA_TMPDIR"],
1448
				envVars["OLLAMA_FLASH_ATTENTION"],
1449
				envVars["OLLAMA_KV_CACHE_TYPE"],
1450
				envVars["OLLAMA_LLM_LIBRARY"],
1451
				envVars["OLLAMA_GPU_OVERHEAD"],
1452
				envVars["OLLAMA_LOAD_TIMEOUT"],
1453
			})
1454
1455
1456
		default:
			appendEnvDocs(cmd, envs)
		}
1457
1458
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1459
1460
	rootCmd.AddCommand(
		serveCmd,
1461
		createCmd,
Patrick Devine's avatar
Patrick Devine committed
1462
		showCmd,
1463
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1464
		stopCmd,
1465
1466
		pullCmd,
		pushCmd,
Patrick Devine's avatar
Patrick Devine committed
1467
		listCmd,
1468
		psCmd,
Patrick Devine's avatar
Patrick Devine committed
1469
		copyCmd,
1470
		deleteCmd,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1471
1472
1473
1474
	)

	return rootCmd
}