cmd.go 34 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
	"github.com/ollama/ollama/progress"
	"github.com/ollama/ollama/server"
42
	"github.com/ollama/ollama/types/model"
43
	"github.com/ollama/ollama/version"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
44
45
)

46
47
48
49
50
51
52
53
54
55
56
57
58
59
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)
60
	if err != nil {
61
		return "", err
62
63
	}

64
	_, err = os.Stat(absName)
65
	if err != nil {
66
		return fn, err
67
	}
68

69
70
71
72
	return absName, nil
}

func CreateHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
73
74
75
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

76
77
78
79
80
81
82
83
84
85
	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
86
		return err
87
88
89
90
91
92
93
94
	} else {
		f, err := os.Open(filename)
		if err != nil {
			return err
		}

		reader = f
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
95
96
	}

97
	modelfile, err := parser.ParseFile(reader)
Michael Yang's avatar
Michael Yang committed
98
99
100
101
102
103
104
105
106
	if err != nil {
		return err
	}

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

107
108
	status := "transferring model data"
	spinner := progress.NewSpinner(status)
109
	p.Add(status, spinner)
110
	defer p.Stop()
111

112
113
114
115
116
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
117
118
	for i := range modelfile.Commands {
		switch modelfile.Commands[i].Name {
Michael Yang's avatar
Michael Yang committed
119
		case "model", "adapter":
Michael Yang's avatar
Michael Yang committed
120
			path := modelfile.Commands[i].Args
Michael Yang's avatar
Michael Yang committed
121
122
123
124
125
126
			if path == "~" {
				path = home
			} else if strings.HasPrefix(path, "~/") {
				path = filepath.Join(home, path[2:])
			}

127
128
129
130
			if !filepath.IsAbs(path) {
				path = filepath.Join(filepath.Dir(filename), path)
			}

131
			fi, err := os.Stat(path)
Michael Yang's avatar
Michael Yang committed
132
			if errors.Is(err, os.ErrNotExist) && modelfile.Commands[i].Name == "model" {
Michael Yang's avatar
Michael Yang committed
133
				continue
Michael Yang's avatar
Michael Yang committed
134
135
136
137
			} else if err != nil {
				return err
			}

138
			if fi.IsDir() {
Michael Yang's avatar
Michael Yang committed
139
140
141
				// this is likely a safetensors or pytorch directory
				// TODO make this work w/ adapters
				tempfile, err := tempZipFiles(path)
142
143
144
				if err != nil {
					return err
				}
Michael Yang's avatar
Michael Yang committed
145
				defer os.RemoveAll(tempfile)
146

Michael Yang's avatar
Michael Yang committed
147
				path = tempfile
Michael Yang's avatar
Michael Yang committed
148
149
			}

150
			digest, err := createBlob(cmd, client, path, spinner)
151
			if err != nil {
Michael Yang's avatar
Michael Yang committed
152
153
154
				return err
			}

Michael Yang's avatar
Michael Yang committed
155
			modelfile.Commands[i].Args = "@" + digest
Michael Yang's avatar
Michael Yang committed
156
157
		}
	}
Michael Yang's avatar
Michael Yang committed
158

Michael Yang's avatar
Michael Yang committed
159
	bars := make(map[string]*progress.Bar)
160
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
161
162
163
164
165
		if resp.Digest != "" {
			spinner.Stop()

			bar, ok := bars[resp.Digest]
			if !ok {
166
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
167
168
169
170
171
172
173
174
175
176
177
178
179
				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)
		}

180
181
182
		return nil
	}

183
	quantize, _ := cmd.Flags().GetString("quantize")
Michael Yang's avatar
Michael Yang committed
184

185
	request := api.CreateRequest{Name: args[0], Modelfile: modelfile.String(), Quantize: quantize}
Michael Yang's avatar
Michael Yang committed
186
	if err := client.Create(cmd.Context(), &request, fn); err != nil {
187
188
189
190
191
192
		return err
	}

	return nil
}

Michael Yang's avatar
Michael Yang committed
193
194
195
196
197
198
199
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
200
201
	detectContentType := func(path string) (string, error) {
		f, err := os.Open(path)
Michael Yang's avatar
Michael Yang committed
202
203
204
		if err != nil {
			return "", err
		}
Michael Yang's avatar
Michael Yang committed
205
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
206

Michael Yang's avatar
Michael Yang committed
207
208
		var b bytes.Buffer
		b.Grow(512)
Michael Yang's avatar
Michael Yang committed
209

Michael Yang's avatar
Michael Yang committed
210
211
212
213
214
215
		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
216
217
	}

Michael Yang's avatar
Michael Yang committed
218
219
220
221
222
223
224
225
226
227
228
	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
229
			}
Michael Yang's avatar
Michael Yang committed
230
231
232
233
234
235
236
237
238
239
		}

		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...)
240
241
242
243
244
245
	} 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
246
247
248
249
	} 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
250
	} else if pt, _ := glob(filepath.Join(path, "consolidated*.pth"), "application/zip"); len(pt) > 0 {
Michael Yang's avatar
Michael Yang committed
251
252
253
254
		// pytorch files might also be unresolved git lfs references; skip if they are
		// covers consolidated.x.pth, consolidated.pth
		files = append(files, pt...)
	} else {
255
		return "", errModelNotFound
Michael Yang's avatar
Michael Yang committed
256
257
258
259
260
261
262
263
264
	}

	// 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
265
266
267
268
269
270
271
272
	// 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
273
274
275
276
277
278
279
280
281
	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...)
	}

282
283
284
	zipfile := zip.NewWriter(tempfile)
	defer zipfile.Close()

Michael Yang's avatar
Michael Yang committed
285
286
287
	for _, file := range files {
		f, err := os.Open(file)
		if err != nil {
Michael Yang's avatar
Michael Yang committed
288
289
			return "", err
		}
Michael Yang's avatar
Michael Yang committed
290
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
291
292
293
294
295
296

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

Michael Yang's avatar
Michael Yang committed
297
		zfi, err := zip.FileInfoHeader(fi)
Michael Yang's avatar
Michael Yang committed
298
299
300
301
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
302
303
304
305
306
		zfi.Name, err = filepath.Rel(path, file)
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
307
		zf, err := zipfile.CreateHeader(zfi)
Michael Yang's avatar
Michael Yang committed
308
309
310
311
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
312
		if _, err := io.Copy(zf, f); err != nil {
Michael Yang's avatar
Michael Yang committed
313
314
315
316
317
318
319
			return "", err
		}
	}

	return tempfile.Name(), nil
}

320
func createBlob(cmd *cobra.Command, client *api.Client, path string, spinner *progress.Spinner) (string, error) {
321
322
323
324
325
326
	bin, err := os.Open(path)
	if err != nil {
		return "", err
	}
	defer bin.Close()

327
328
329
330
331
332
333
	// Get file info to retrieve the size
	fileInfo, err := bin.Stat()
	if err != nil {
		return "", err
	}
	fileSize := fileInfo.Size()

334
335
336
337
	hash := sha256.New()
	if _, err := io.Copy(hash, bin); err != nil {
		return "", err
	}
338
339
340
341

	if _, err := bin.Seek(0, io.SeekStart); err != nil {
		return "", err
	}
342

343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
	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
			}
		}
	}()

364
	digest := fmt.Sprintf("sha256:%x", hash.Sum(nil))
365
	if err = client.CreateBlob(cmd.Context(), digest, io.TeeReader(bin, &pw)); err != nil {
366
367
368
369
370
		return "", err
	}
	return digest, nil
}

371
372
373
374
375
376
377
378
379
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
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
412
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
}

413
func RunHandler(cmd *cobra.Command, args []string) error {
414
415
416
	interactive := true

	opts := runOptions{
417
418
419
		Model:    args[0],
		WordWrap: os.Getenv("TERM") == "xterm-256color",
		Options:  map[string]interface{}{},
420
421
422
423
424
425
426
427
	}

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

428
429
430
431
432
433
434
435
436
437
438
439
	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}
	}

440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
	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
	}
456
457
458
459
	// Be quiet if we're redirecting to a pipe or file
	if !term.IsTerminal(int(os.Stdout.Fd())) {
		interactive = false
	}
460
461
462
463
464
465
466

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

467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
	// 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
489
490
	}

491
	opts.MultiModal = len(info.ProjectorInfo) != 0
492
493
494
	opts.ParentModel = info.Details.ParentModel

	if interactive {
Patrick Devine's avatar
Patrick Devine committed
495
		if err := loadOrUnloadModel(cmd, &opts); err != nil {
Michael Yang's avatar
Michael Yang committed
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
			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()
			}
		}

511
512
513
		return generateInteractive(cmd, opts)
	}
	return generate(cmd, opts)
Bruce MacDonald's avatar
Bruce MacDonald committed
514
515
}

516
func PushHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
517
	client, err := api.ClientFromEnvironment()
518
519
520
	if err != nil {
		return err
	}
521

522
523
524
525
526
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
527
528
529
530
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)
531
532
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
533

534
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
535
		if resp.Digest != "" {
536
537
538
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
539
540
541

			bar, ok := bars[resp.Digest]
			if !ok {
542
				bar = progress.NewBar(fmt.Sprintf("pushing %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
543
544
545
546
547
548
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
549
550
551
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
552
553
554
555
556
557

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

558
559
560
		return nil
	}

Michael Yang's avatar
Michael Yang committed
561
	request := api.PushRequest{Name: args[0], Insecure: insecure}
562
563

	n := model.ParseName(args[0])
Michael Yang's avatar
Michael Yang committed
564
	if err := client.Push(cmd.Context(), &request, fn); err != nil {
565
566
567
568
569
570
		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
571
572
573
		return err
	}

574
	p.Stop()
575
	spinner.Stop()
576
577
578
579
580
581
582
583

	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
584
	return nil
585
586
}

587
func ListHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
588
	client, err := api.ClientFromEnvironment()
589
590
591
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
592

Michael Yang's avatar
Michael Yang committed
593
	models, err := client.List(cmd.Context())
Patrick Devine's avatar
Patrick Devine committed
594
595
596
597
598
599
600
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
Michael Yang's avatar
Michael Yang committed
601
		if len(args) == 0 || strings.HasPrefix(m.Name, args[0]) {
602
			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
603
		}
Patrick Devine's avatar
Patrick Devine committed
604
605
606
	}

	table := tablewriter.NewWriter(os.Stdout)
Patrick Devine's avatar
Patrick Devine committed
607
	table.SetHeader([]string{"NAME", "ID", "SIZE", "MODIFIED"})
Patrick Devine's avatar
Patrick Devine committed
608
609
610
611
612
	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
613
	table.SetTablePadding("    ")
Patrick Devine's avatar
Patrick Devine committed
614
615
616
617
618
619
	table.AppendBulk(data)
	table.Render()

	return nil
}

620
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
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
648
649
650
651
652
653
654
655
656

			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})
657
658
659
660
661
662
663
664
665
666
		}
	}

	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
667
	table.SetTablePadding("    ")
668
669
670
671
672
673
	table.AppendBulk(data)
	table.Render()

	return nil
}

674
func DeleteHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
675
	client, err := api.ClientFromEnvironment()
676
677
678
	if err != nil {
		return err
	}
679

680
681
682
683
684
685
686
687
688
689
690
	// 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)
		}
	}

691
692
	for _, name := range args {
		req := api.DeleteRequest{Name: name}
Michael Yang's avatar
Michael Yang committed
693
		if err := client.Delete(cmd.Context(), &req); err != nil {
694
695
696
			return err
		}
		fmt.Printf("deleted '%s'\n", name)
697
698
699
700
	}
	return nil
}

Patrick Devine's avatar
Patrick Devine committed
701
func ShowHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
702
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
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
736
737
738
739
740
741
742
743
744
745
746
747
	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 {
748
		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
749
750
	}

751
752
753
754
755
	req := api.ShowRequest{Name: args[0]}
	resp, err := client.Show(cmd.Context(), &req)
	if err != nil {
		return err
	}
756

757
	if flagsSet == 1 {
758
759
760
761
762
763
764
765
		switch showType {
		case "license":
			fmt.Println(resp.License)
		case "modelfile":
			fmt.Println(resp.Modelfile)
		case "parameters":
			fmt.Println(resp.Parameters)
		case "system":
766
			fmt.Print(resp.System)
767
		case "template":
768
			fmt.Print(resp.Template)
769
770
771
		}

		return nil
Patrick Devine's avatar
Patrick Devine committed
772
773
	}

Michael Yang's avatar
Michael Yang committed
774
	return showInfo(resp, os.Stdout)
775
776
}

Michael Yang's avatar
Michael Yang committed
777
778
779
780
781
782
783
784
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("    ")
785

Michael Yang's avatar
Michael Yang committed
786
787
788
		switch header {
		case "Template", "System", "License":
			table.SetColWidth(100)
789
790
		}

Michael Yang's avatar
Michael Yang committed
791
792
793
		table.AppendBulk(rows())
		table.Render()
		fmt.Fprintln(w)
Patrick Devine's avatar
Patrick Devine committed
794
795
	}

Michael Yang's avatar
Michael Yang committed
796
797
798
799
800
801
802
803
804
805
806
807
808
809
	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
	})
810

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

Michael Yang's avatar
Michael Yang committed
822
823
824
825
826
827
828
829
830
831
	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
		})
832
833
	}

Michael Yang's avatar
Michael Yang committed
834
835
836
837
838
	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)})
839
840
			}
		}
Michael Yang's avatar
Michael Yang committed
841
		return
842
843
	}

Michael Yang's avatar
Michael Yang committed
844
845
846
847
848
	if resp.System != "" {
		tableRender("System", func() [][]string {
			return head(resp.System, 2)
		})
	}
849

Michael Yang's avatar
Michael Yang committed
850
851
852
853
	if resp.License != "" {
		tableRender("License", func() [][]string {
			return head(resp.License, 2)
		})
854
	}
Michael Yang's avatar
Michael Yang committed
855
856

	return nil
857
858
}

Patrick Devine's avatar
Patrick Devine committed
859
func CopyHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
860
	client, err := api.ClientFromEnvironment()
861
862
863
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
864
865

	req := api.CopyRequest{Source: args[0], Destination: args[1]}
Michael Yang's avatar
Michael Yang committed
866
	if err := client.Copy(cmd.Context(), &req); err != nil {
Patrick Devine's avatar
Patrick Devine committed
867
868
869
870
871
872
		return err
	}
	fmt.Printf("copied '%s' to '%s'\n", args[0], args[1])
	return nil
}

873
func PullHandler(cmd *cobra.Command, args []string) error {
874
875
876
877
878
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
879
	client, err := api.ClientFromEnvironment()
880
881
882
	if err != nil {
		return err
	}
883

Michael Yang's avatar
Michael Yang committed
884
885
886
887
888
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

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

889
890
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
891

892
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
893
		if resp.Digest != "" {
894
895
896
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
897
898
899

			bar, ok := bars[resp.Digest]
			if !ok {
900
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
901
902
903
904
905
906
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
907
908
909
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
910
911
912
913
914
915

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

916
917
		return nil
	}
918

Michael Yang's avatar
Michael Yang committed
919
	request := api.PullRequest{Name: args[0], Insecure: insecure}
Michael Yang's avatar
Michael Yang committed
920
	if err := client.Pull(cmd.Context(), &request, fn); err != nil {
Michael Yang's avatar
Michael Yang committed
921
922
923
924
		return err
	}

	return nil
Michael Yang's avatar
Michael Yang committed
925
926
}

927
928
type generateContextKey string

929
type runOptions struct {
930
931
932
933
934
935
936
937
938
939
	Model       string
	ParentModel string
	Prompt      string
	Messages    []api.Message
	WordWrap    bool
	Format      string
	System      string
	Images      []api.ImageData
	Options     map[string]interface{}
	MultiModal  bool
940
	KeepAlive   *api.Duration
941
942
}

943
944
945
946
947
948
949
950
951
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
952
953
			if state.lineLength+1 > termWidth-5 {
				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
954
955
956
957
958
959
960
					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
961
962
				a := runewidth.StringWidth(state.wordBuffer)
				if a > 0 {
963
					fmt.Printf("\x1b[%dD", a)
964
965
				}
				fmt.Printf("\x1b[K\n")
966
				fmt.Printf("%s%c", state.wordBuffer, ch)
967
968
969
				chWidth := runewidth.RuneWidth(ch)

				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
970
971
			} else {
				fmt.Print(string(ch))
972
973
974
975
				state.lineLength += runewidth.RuneWidth(ch)
				if runewidth.RuneWidth(ch) >= 2 {
					state.wordBuffer = ""
					continue
Josh Yan's avatar
Josh Yan committed
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
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044

				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,
	}

1045
1046
1047
1048
	if opts.KeepAlive != nil {
		req.KeepAlive = opts.KeepAlive
	}

1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
	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
1074
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
1075
	if err != nil {
1076
		return err
Patrick Devine's avatar
Patrick Devine committed
1077
	}
Michael Yang's avatar
Michael Yang committed
1078

Michael Yang's avatar
Michael Yang committed
1079
	p := progress.NewProgress(os.Stderr)
1080
	defer p.StopAndClear()
1081

Michael Yang's avatar
Michael Yang committed
1082
1083
1084
	spinner := progress.NewSpinner("")
	p.Add("", spinner)

1085
1086
1087
1088
1089
1090
1091
	var latest api.GenerateResponse

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

Michael Yang's avatar
Michael Yang committed
1092
	ctx, cancel := context.WithCancel(cmd.Context())
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
	defer cancel()

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

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

1103
	var state *displayResponseState = &displayResponseState{}
1104

1105
	fn := func(response api.GenerateResponse) error {
Michael Yang's avatar
Michael Yang committed
1106
		p.StopAndClear()
1107

Patrick Devine's avatar
Patrick Devine committed
1108
		latest = response
1109
		content := response.Response
1110

1111
		displayResponse(content, opts.WordWrap, state)
1112

Patrick Devine's avatar
Patrick Devine committed
1113
1114
		return nil
	}
1115

1116
1117
1118
1119
1120
1121
1122
	if opts.MultiModal {
		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
		if err != nil {
			return err
		}
	}

Michael Yang's avatar
Michael Yang committed
1123
	request := api.GenerateRequest{
1124
1125
1126
1127
1128
1129
1130
1131
		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
1132
1133
1134
	}

	if err := client.Generate(ctx, &request, fn); err != nil {
1135
		if errors.Is(err, context.Canceled) {
1136
			return nil
1137
		}
1138
		return err
Patrick Devine's avatar
Patrick Devine committed
1139
	}
1140

1141
	if opts.Prompt != "" {
Michael Yang's avatar
Michael Yang committed
1142
1143
		fmt.Println()
		fmt.Println()
Patrick Devine's avatar
Patrick Devine committed
1144
	}
1145

1146
1147
1148
1149
	if !latest.Done {
		return nil
	}

Patrick Devine's avatar
Patrick Devine committed
1150
1151
	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
1152
		return err
Patrick Devine's avatar
Patrick Devine committed
1153
	}
Michael Yang's avatar
Michael Yang committed
1154

Patrick Devine's avatar
Patrick Devine committed
1155
1156
	if verbose {
		latest.Summary()
Michael Yang's avatar
Michael Yang committed
1157
	}
Michael Yang's avatar
Michael Yang committed
1158

Patrick Devine's avatar
Patrick Devine committed
1159
1160
1161
	ctx = context.WithValue(cmd.Context(), generateContextKey("context"), latest.Context)
	cmd.SetContext(ctx)

1162
	return nil
Michael Yang's avatar
Michael Yang committed
1163
1164
}

1165
func RunServer(_ *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1166
	if err := initializeKeypair(); err != nil {
1167
1168
1169
		return err
	}

Michael Yang's avatar
host  
Michael Yang committed
1170
	ln, err := net.Listen("tcp", envconfig.Host().Host)
1171
1172
1173
	if err != nil {
		return err
	}
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1174

1175
1176
1177
1178
1179
1180
	err = server.Serve(ln)
	if errors.Is(err, http.ErrServerClosed) {
		return nil
	}

	return err
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1181
1182
}

1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
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
1195
		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
1196
1197
1198
1199
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1200
		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
1201
1202
1203
1204
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1205
		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
1206
1207
1208
			return fmt.Errorf("could not create directory %w", err)
		}

Michael Yang's avatar
Michael Yang committed
1209
		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
1210
1211
1212
			return err
		}

Michael Yang's avatar
Michael Yang committed
1213
		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
1214
1215
1216
1217
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1218
		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
1219

Michael Yang's avatar
Michael Yang committed
1220
		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
1221
1222
1223
			return err
		}

Michael Yang's avatar
Michael Yang committed
1224
		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
1225
1226
1227
1228
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1229
func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1230
	client, err := api.ClientFromEnvironment()
1231
1232
1233
	if err != nil {
		return err
	}
Michael Yang's avatar
Michael Yang committed
1234
	if err := client.Heartbeat(cmd.Context()); err != nil {
1235
		if !strings.Contains(err.Error(), " refused") {
Bruce MacDonald's avatar
Bruce MacDonald committed
1236
1237
			return err
		}
1238
		if err := startApp(cmd.Context(), client); err != nil {
Michael Yang's avatar
lint  
Michael Yang committed
1239
			return errors.New("could not connect to ollama app, is it running?")
1240
1241
1242
1243
1244
		}
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1245
1246
1247
1248
1249
1250
1251
1252
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
1253
1254
1255
1256
1257
		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
1258
1259
	}

1260
	if serverVersion != version.Version {
Michael Yang's avatar
Michael Yang committed
1261
		fmt.Printf("Warning: client version is %s\n", version.Version)
1262
	}
Michael Yang's avatar
Michael Yang committed
1263
1264
}

1265
func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
1266
1267
1268
1269
1270
	if len(envs) == 0 {
		return
	}

	envUsage := `
1271
1272
Environment Variables:
`
1273
	for _, e := range envs {
1274
		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
1275
1276
1277
	}

	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
1278
1279
}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1280
1281
func NewCLI() *cobra.Command {
	log.SetFlags(log.LstdFlags | log.Lshortfile)
Michael Yang's avatar
Michael Yang committed
1282
	cobra.EnableCommandSorting = false
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1283

1284
	if runtime.GOOS == "windows" && term.IsTerminal(int(os.Stdout.Fd())) {
1285
		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
1286
1287
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1288
	rootCmd := &cobra.Command{
1289
1290
1291
1292
		Use:           "ollama",
		Short:         "Large language model runner",
		SilenceUsage:  true,
		SilenceErrors: true,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1293
1294
1295
		CompletionOptions: cobra.CompletionOptions{
			DisableDefaultCmd: true,
		},
Michael Yang's avatar
Michael Yang committed
1296
1297
1298
1299
1300
1301
1302
1303
		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
1304
1305
	}

Michael Yang's avatar
Michael Yang committed
1306
	rootCmd.Flags().BoolP("version", "v", false, "Show version information")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1307

1308
	createCmd := &cobra.Command{
1309
1310
		Use:     "create MODEL",
		Short:   "Create a model from a Modelfile",
Michael Yang's avatar
Michael Yang committed
1311
		Args:    cobra.ExactArgs(1),
1312
1313
		PreRunE: checkServerHeartbeat,
		RunE:    CreateHandler,
1314
1315
	}

1316
	createCmd.Flags().StringP("file", "f", "", "Name of the Modelfile (default \"Modelfile\"")
1317
	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_0)")
1318

Patrick Devine's avatar
Patrick Devine committed
1319
1320
1321
	showCmd := &cobra.Command{
		Use:     "show MODEL",
		Short:   "Show information for a model",
Michael Yang's avatar
Michael Yang committed
1322
		Args:    cobra.ExactArgs(1),
Patrick Devine's avatar
Patrick Devine committed
1323
1324
1325
1326
1327
1328
1329
1330
		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")
1331
	showCmd.Flags().Bool("system", false, "Show system message of a model")
Patrick Devine's avatar
Patrick Devine committed
1332

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1333
	runCmd := &cobra.Command{
1334
1335
1336
1337
1338
		Use:     "run MODEL [PROMPT]",
		Short:   "Run a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    RunHandler,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1339
1340
	}

1341
	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
1342
	runCmd.Flags().Bool("verbose", false, "Show timings for response")
1343
	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
1344
	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1345
	runCmd.Flags().String("format", "", "Response format (e.g. json)")
Patrick Devine's avatar
Patrick Devine committed
1346
1347
1348
1349
1350
1351
1352
1353
1354

	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
1355
1356
1357
1358
	serveCmd := &cobra.Command{
		Use:     "serve",
		Aliases: []string{"start"},
		Short:   "Start ollama",
Michael Yang's avatar
Michael Yang committed
1359
		Args:    cobra.ExactArgs(0),
Michael Yang's avatar
Michael Yang committed
1360
		RunE:    RunServer,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1361
1362
	}

1363
	pullCmd := &cobra.Command{
1364
1365
		Use:     "pull MODEL",
		Short:   "Pull a model from a registry",
Michael Yang's avatar
Michael Yang committed
1366
		Args:    cobra.ExactArgs(1),
1367
1368
		PreRunE: checkServerHeartbeat,
		RunE:    PullHandler,
1369
1370
	}

1371
1372
	pullCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1373
	pushCmd := &cobra.Command{
1374
1375
		Use:     "push MODEL",
		Short:   "Push a model to a registry",
Michael Yang's avatar
Michael Yang committed
1376
		Args:    cobra.ExactArgs(1),
1377
1378
		PreRunE: checkServerHeartbeat,
		RunE:    PushHandler,
1379
1380
	}

1381
1382
	pushCmd.Flags().Bool("insecure", false, "Use an insecure registry")

Patrick Devine's avatar
Patrick Devine committed
1383
	listCmd := &cobra.Command{
1384
		Use:     "list",
Patrick Devine's avatar
Patrick Devine committed
1385
		Aliases: []string{"ls"},
1386
		Short:   "List models",
1387
		PreRunE: checkServerHeartbeat,
1388
		RunE:    ListHandler,
1389
	}
1390
1391
1392
1393
1394
1395
1396
1397

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

Patrick Devine's avatar
Patrick Devine committed
1398
	copyCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1399
		Use:     "cp SOURCE DESTINATION",
1400
		Short:   "Copy a model",
Michael Yang's avatar
Michael Yang committed
1401
		Args:    cobra.ExactArgs(2),
1402
1403
		PreRunE: checkServerHeartbeat,
		RunE:    CopyHandler,
Patrick Devine's avatar
Patrick Devine committed
1404
1405
	}

1406
	deleteCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1407
		Use:     "rm MODEL [MODEL...]",
1408
1409
1410
1411
		Short:   "Remove a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    DeleteHandler,
Patrick Devine's avatar
Patrick Devine committed
1412
1413
	}

1414
1415
1416
	envVars := envconfig.AsMap()

	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
1417

1418
1419
1420
1421
	for _, cmd := range []*cobra.Command{
		createCmd,
		showCmd,
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1422
		stopCmd,
1423
1424
1425
		pullCmd,
		pushCmd,
		listCmd,
1426
		psCmd,
1427
1428
		copyCmd,
		deleteCmd,
1429
		serveCmd,
1430
	} {
1431
1432
		switch cmd {
		case runCmd:
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
			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"],
1445
				envVars["OLLAMA_SCHED_SPREAD"],
1446
				envVars["OLLAMA_TMPDIR"],
1447
1448
				envVars["OLLAMA_FLASH_ATTENTION"],
				envVars["OLLAMA_LLM_LIBRARY"],
1449
				envVars["OLLAMA_GPU_OVERHEAD"],
1450
				envVars["OLLAMA_LOAD_TIMEOUT"],
1451
			})
1452
1453
1454
		default:
			appendEnvDocs(cmd, envs)
		}
1455
1456
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1457
1458
	rootCmd.AddCommand(
		serveCmd,
1459
		createCmd,
Patrick Devine's avatar
Patrick Devine committed
1460
		showCmd,
1461
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1462
		stopCmd,
1463
1464
		pullCmd,
		pushCmd,
Patrick Devine's avatar
Patrick Devine committed
1465
		listCmd,
1466
		psCmd,
Patrick Devine's avatar
Patrick Devine committed
1467
		copyCmd,
1468
		deleteCmd,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1469
1470
1471
1472
	)

	return rootCmd
}