cmd.go 30.7 KB
Newer Older
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1
2
3
package cmd

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

27
	"github.com/containerd/console"
28
	"github.com/mattn/go-runewidth"
Patrick Devine's avatar
Patrick Devine committed
29
	"github.com/olekukonko/tablewriter"
Michael Yang's avatar
Michael Yang committed
30
	"github.com/spf13/cobra"
31
	"golang.org/x/crypto/ssh"
32
	"golang.org/x/term"
Michael Yang's avatar
Michael Yang committed
33

34
	"github.com/ollama/ollama/api"
35
	"github.com/ollama/ollama/envconfig"
36
	"github.com/ollama/ollama/format"
37
	"github.com/ollama/ollama/llama"
38
	"github.com/ollama/ollama/parser"
39
	"github.com/ollama/ollama/progress"
Jesse Gross's avatar
Jesse Gross committed
40
	"github.com/ollama/ollama/runner"
41
	"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
var errModelfileNotFound = errors.New("specified Modelfile wasn't found")
47
48

func getModelfileName(cmd *cobra.Command) (string, error) {
49
	filename, _ := cmd.Flags().GetString("file")
50
51
52
53
54
55

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

	absName, err := filepath.Abs(filename)
56
	if err != nil {
57
		return "", err
58
59
	}

60
	_, err = os.Stat(absName)
61
	if err != nil {
62
		return "", err
63
	}
64

65
66
67
68
	return absName, nil
}

func CreateHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
69
70
71
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

72
73
74
75
76
77
78
79
80
81
	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
82
		return err
83
84
85
86
87
88
89
90
	} else {
		f, err := os.Open(filename)
		if err != nil {
			return err
		}

		reader = f
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
91
92
	}

93
	modelfile, err := parser.ParseFile(reader)
Michael Yang's avatar
Michael Yang committed
94
95
96
97
	if err != nil {
		return err
	}

98
99
100
101
	status := "gathering model components"
	spinner := progress.NewSpinner(status)
	p.Add(status, spinner)

102
	req, err := modelfile.CreateRequest(filepath.Dir(filename))
Michael Yang's avatar
Michael Yang committed
103
104
105
	if err != nil {
		return err
	}
106
	spinner.Stop()
Michael Yang's avatar
Michael Yang committed
107

108
109
110
111
112
	req.Name = args[0]
	quantize, _ := cmd.Flags().GetString("quantize")
	if quantize != "" {
		req.Quantize = quantize
	}
113

114
115
116
117
118
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

119
120
121
122
	if len(req.Files) > 0 {
		fileMap := map[string]string{}
		for f, digest := range req.Files {
			if _, err := createBlob(cmd, client, f, digest, p); err != nil {
Michael Yang's avatar
Michael Yang committed
123
124
				return err
			}
125
126
127
128
			fileMap[filepath.Base(f)] = digest
		}
		req.Files = fileMap
	}
Michael Yang's avatar
Michael Yang committed
129

130
131
132
133
	if len(req.Adapters) > 0 {
		fileMap := map[string]string{}
		for f, digest := range req.Adapters {
			if _, err := createBlob(cmd, client, f, digest, p); err != nil {
Michael Yang's avatar
Michael Yang committed
134
135
				return err
			}
136
			fileMap[filepath.Base(f)] = digest
Michael Yang's avatar
Michael Yang committed
137
		}
138
		req.Adapters = fileMap
Michael Yang's avatar
Michael Yang committed
139
	}
Michael Yang's avatar
Michael Yang committed
140

Michael Yang's avatar
Michael Yang committed
141
	bars := make(map[string]*progress.Bar)
142
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
143
144
145
		if resp.Digest != "" {
			bar, ok := bars[resp.Digest]
			if !ok {
146
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
147
148
149
150
151
152
153
154
155
156
157
158
159
				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)
		}

160
161
162
		return nil
	}

163
	if err := client.Create(cmd.Context(), req, fn); err != nil {
164
165
166
		if strings.Contains(err.Error(), "path or Modelfile are required") {
			return fmt.Errorf("the ollama server must be updated to use `ollama create` with this client")
		}
167
168
169
170
171
172
		return err
	}

	return nil
}

173
174
func createBlob(cmd *cobra.Command, client *api.Client, path string, digest string, p *progress.Progress) (string, error) {
	realPath, err := filepath.EvalSymlinks(path)
Michael Yang's avatar
Michael Yang committed
175
176
177
178
	if err != nil {
		return "", err
	}

179
	bin, err := os.Open(realPath)
180
181
182
183
184
	if err != nil {
		return "", err
	}
	defer bin.Close()

185
186
187
188
189
190
191
192
	// Get file info to retrieve the size
	fileInfo, err := bin.Stat()
	if err != nil {
		return "", err
	}
	fileSize := fileInfo.Size()

	var pw progressWriter
193
194
195
196
	status := fmt.Sprintf("copying file %s 0%%", digest)
	spinner := progress.NewSpinner(status)
	p.Add(status, spinner)
	defer spinner.Stop()
197
198
199
200
201
202
203
204
205
206

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

	go func() {
		ticker := time.NewTicker(60 * time.Millisecond)
		defer ticker.Stop()
		for {
			select {
			case <-ticker.C:
207
				spinner.SetMessage(fmt.Sprintf("copying file %s %d%%", digest, int(100*pw.n.Load()/fileSize)))
208
			case <-done:
209
				spinner.SetMessage(fmt.Sprintf("copying file %s 100%%", digest))
210
211
212
213
214
215
				return
			}
		}
	}()

	if err = client.CreateBlob(cmd.Context(), digest, io.TeeReader(bin, &pw)); err != nil {
216
217
218
219
220
		return "", err
	}
	return digest, nil
}

221
222
223
224
225
226
227
228
229
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
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
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])
		}
259
		return err
Patrick Devine's avatar
Patrick Devine committed
260
261
262
263
	}
	return nil
}

264
func RunHandler(cmd *cobra.Command, args []string) error {
265
266
267
	interactive := true

	opts := runOptions{
268
269
270
		Model:    args[0],
		WordWrap: os.Getenv("TERM") == "xterm-256color",
		Options:  map[string]interface{}{},
271
272
273
274
275
276
277
278
	}

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

279
280
281
282
283
284
285
286
287
288
289
290
	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}
	}

291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
	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
	}
307
308
309
310
	// Be quiet if we're redirecting to a pipe or file
	if !term.IsTerminal(int(os.Stdout.Fd())) {
		interactive = false
	}
311
312
313
314
315
316
317

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

318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
	// 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
340
341
	}

342
343
344
345
346
347
348
349
350
351
	if len(info.ProjectorInfo) != 0 {
		opts.MultiModal = true
	}
	for k := range info.ModelInfo {
		if strings.Contains(k, ".vision.") {
			opts.MultiModal = true
			break
		}
	}

352
353
354
	opts.ParentModel = info.Details.ParentModel

	if interactive {
Patrick Devine's avatar
Patrick Devine committed
355
		if err := loadOrUnloadModel(cmd, &opts); err != nil {
Michael Yang's avatar
Michael Yang committed
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
			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()
			}
		}

371
372
373
		return generateInteractive(cmd, opts)
	}
	return generate(cmd, opts)
Bruce MacDonald's avatar
Bruce MacDonald committed
374
375
}

376
func PushHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
377
	client, err := api.ClientFromEnvironment()
378
379
380
	if err != nil {
		return err
	}
381

382
383
384
385
386
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
387
388
389
390
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)
391
392
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
393

394
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
395
		if resp.Digest != "" {
396
397
398
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
399
400
401

			bar, ok := bars[resp.Digest]
			if !ok {
402
				bar = progress.NewBar(fmt.Sprintf("pushing %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
403
404
405
406
407
408
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
409
410
411
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
412
413
414
415
416
417

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

418
419
420
		return nil
	}

Michael Yang's avatar
Michael Yang committed
421
	request := api.PushRequest{Name: args[0], Insecure: insecure}
422
423

	n := model.ParseName(args[0])
Michael Yang's avatar
Michael Yang committed
424
	if err := client.Push(cmd.Context(), &request, fn); err != nil {
425
426
427
428
429
430
		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
431
432
433
		return err
	}

434
	p.Stop()
435
	spinner.Stop()
436
437
438
439
440
441
442
443

	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
444
	return nil
445
446
}

447
func ListHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
448
	client, err := api.ClientFromEnvironment()
449
450
451
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
452

Michael Yang's avatar
Michael Yang committed
453
	models, err := client.List(cmd.Context())
Patrick Devine's avatar
Patrick Devine committed
454
455
456
457
458
459
460
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
461
		if len(args) == 0 || strings.HasPrefix(strings.ToLower(m.Name), strings.ToLower(args[0])) {
462
			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
463
		}
Patrick Devine's avatar
Patrick Devine committed
464
465
466
	}

	table := tablewriter.NewWriter(os.Stdout)
Patrick Devine's avatar
Patrick Devine committed
467
	table.SetHeader([]string{"NAME", "ID", "SIZE", "MODIFIED"})
Patrick Devine's avatar
Patrick Devine committed
468
469
470
471
472
	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
473
	table.SetTablePadding("    ")
Patrick Devine's avatar
Patrick Devine committed
474
475
476
477
478
479
	table.AppendBulk(data)
	table.Render()

	return nil
}

480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
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
508
509
510
511
512
513
514
515
516

			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})
517
518
519
520
521
522
523
524
525
526
		}
	}

	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
527
	table.SetTablePadding("    ")
528
529
530
531
532
533
	table.AppendBulk(data)
	table.Render()

	return nil
}

534
func DeleteHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
535
	client, err := api.ClientFromEnvironment()
536
537
538
	if err != nil {
		return err
	}
539

540
541
542
543
544
545
546
547
548
549
550
	// 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)
		}
	}

551
552
	for _, name := range args {
		req := api.DeleteRequest{Name: name}
Michael Yang's avatar
Michael Yang committed
553
		if err := client.Delete(cmd.Context(), &req); err != nil {
554
555
556
			return err
		}
		fmt.Printf("deleted '%s'\n", name)
557
558
559
560
	}
	return nil
}

Patrick Devine's avatar
Patrick Devine committed
561
func ShowHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
562
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
	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 {
608
		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
609
610
	}

611
612
613
614
615
	req := api.ShowRequest{Name: args[0]}
	resp, err := client.Show(cmd.Context(), &req)
	if err != nil {
		return err
	}
616

617
	if flagsSet == 1 {
618
619
620
621
622
623
624
625
		switch showType {
		case "license":
			fmt.Println(resp.License)
		case "modelfile":
			fmt.Println(resp.Modelfile)
		case "parameters":
			fmt.Println(resp.Parameters)
		case "system":
626
			fmt.Print(resp.System)
627
		case "template":
628
			fmt.Print(resp.Template)
629
630
631
		}

		return nil
Patrick Devine's avatar
Patrick Devine committed
632
633
	}

Michael Yang's avatar
Michael Yang committed
634
	return showInfo(resp, os.Stdout)
635
636
}

Michael Yang's avatar
Michael Yang committed
637
638
639
640
641
642
643
644
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("    ")
645

Michael Yang's avatar
Michael Yang committed
646
647
648
		switch header {
		case "Template", "System", "License":
			table.SetColWidth(100)
649
650
		}

Michael Yang's avatar
Michael Yang committed
651
652
653
		table.AppendBulk(rows())
		table.Render()
		fmt.Fprintln(w)
Patrick Devine's avatar
Patrick Devine committed
654
655
	}

Michael Yang's avatar
Michael Yang committed
656
657
658
659
660
661
662
663
664
665
666
667
668
669
	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
	})
670

Michael Yang's avatar
Michael Yang committed
671
672
673
674
675
676
677
678
679
	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
		})
680
681
	}

Michael Yang's avatar
Michael Yang committed
682
683
684
685
686
687
688
689
690
691
	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
		})
692
693
	}

Michael Yang's avatar
Michael Yang committed
694
695
696
697
698
	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)})
699
700
			}
		}
Michael Yang's avatar
Michael Yang committed
701
		return
702
703
	}

Michael Yang's avatar
Michael Yang committed
704
705
706
707
708
	if resp.System != "" {
		tableRender("System", func() [][]string {
			return head(resp.System, 2)
		})
	}
709

Michael Yang's avatar
Michael Yang committed
710
711
712
713
	if resp.License != "" {
		tableRender("License", func() [][]string {
			return head(resp.License, 2)
		})
714
	}
Michael Yang's avatar
Michael Yang committed
715
716

	return nil
717
718
}

Patrick Devine's avatar
Patrick Devine committed
719
func CopyHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
720
	client, err := api.ClientFromEnvironment()
721
722
723
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
724
725

	req := api.CopyRequest{Source: args[0], Destination: args[1]}
Michael Yang's avatar
Michael Yang committed
726
	if err := client.Copy(cmd.Context(), &req); err != nil {
Patrick Devine's avatar
Patrick Devine committed
727
728
729
730
731
732
		return err
	}
	fmt.Printf("copied '%s' to '%s'\n", args[0], args[1])
	return nil
}

733
func PullHandler(cmd *cobra.Command, args []string) error {
734
735
736
737
738
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
739
	client, err := api.ClientFromEnvironment()
740
741
742
	if err != nil {
		return err
	}
743

Michael Yang's avatar
Michael Yang committed
744
745
746
747
748
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

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

749
750
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
751

752
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
753
		if resp.Digest != "" {
754
755
756
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
757
758
759

			bar, ok := bars[resp.Digest]
			if !ok {
760
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
761
762
763
764
765
766
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
767
768
769
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
770
771
772
773
774
775

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

776
777
		return nil
	}
778

Michael Yang's avatar
Michael Yang committed
779
	request := api.PullRequest{Name: args[0], Insecure: insecure}
Michael Yang's avatar
Michael Yang committed
780
	if err := client.Pull(cmd.Context(), &request, fn); err != nil {
Michael Yang's avatar
Michael Yang committed
781
782
783
784
		return err
	}

	return nil
Michael Yang's avatar
Michael Yang committed
785
786
}

787
788
type generateContextKey string

789
type runOptions struct {
790
791
792
793
794
795
796
797
798
799
	Model       string
	ParentModel string
	Prompt      string
	Messages    []api.Message
	WordWrap    bool
	Format      string
	System      string
	Images      []api.ImageData
	Options     map[string]interface{}
	MultiModal  bool
800
	KeepAlive   *api.Duration
801
802
}

803
804
805
806
807
808
809
810
811
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
812
813
			if state.lineLength+1 > termWidth-5 {
				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
814
815
816
817
818
819
820
					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
821
822
				a := runewidth.StringWidth(state.wordBuffer)
				if a > 0 {
823
					fmt.Printf("\x1b[%dD", a)
824
825
				}
				fmt.Printf("\x1b[K\n")
826
				fmt.Printf("%s%c", state.wordBuffer, ch)
827
828
829
				chWidth := runewidth.RuneWidth(ch)

				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
830
831
			} else {
				fmt.Print(string(ch))
832
833
834
835
				state.lineLength += runewidth.RuneWidth(ch)
				if runewidth.RuneWidth(ch) >= 2 {
					state.wordBuffer = ""
					continue
Josh Yan's avatar
Josh Yan committed
836
				}
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897

				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
	}

898
899
900
901
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

902
903
904
	req := &api.ChatRequest{
		Model:    opts.Model,
		Messages: opts.Messages,
905
		Format:   json.RawMessage(opts.Format),
906
907
908
		Options:  opts.Options,
	}

909
910
911
912
	if opts.KeepAlive != nil {
		req.KeepAlive = opts.KeepAlive
	}

913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
	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
938
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
939
	if err != nil {
940
		return err
Patrick Devine's avatar
Patrick Devine committed
941
	}
Michael Yang's avatar
Michael Yang committed
942

Michael Yang's avatar
Michael Yang committed
943
	p := progress.NewProgress(os.Stderr)
944
	defer p.StopAndClear()
945

Michael Yang's avatar
Michael Yang committed
946
947
948
	spinner := progress.NewSpinner("")
	p.Add("", spinner)

949
950
951
952
953
954
955
	var latest api.GenerateResponse

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

Michael Yang's avatar
Michael Yang committed
956
	ctx, cancel := context.WithCancel(cmd.Context())
957
958
959
960
961
962
963
964
965
966
	defer cancel()

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

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

967
	var state *displayResponseState = &displayResponseState{}
968

969
	fn := func(response api.GenerateResponse) error {
Michael Yang's avatar
Michael Yang committed
970
		p.StopAndClear()
971

Patrick Devine's avatar
Patrick Devine committed
972
		latest = response
973
		content := response.Response
974

975
		displayResponse(content, opts.WordWrap, state)
976

Patrick Devine's avatar
Patrick Devine committed
977
978
		return nil
	}
979

980
981
982
983
984
985
986
	if opts.MultiModal {
		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
		if err != nil {
			return err
		}
	}

987
988
989
990
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

Michael Yang's avatar
Michael Yang committed
991
	request := api.GenerateRequest{
992
993
994
995
		Model:     opts.Model,
		Prompt:    opts.Prompt,
		Context:   generateContext,
		Images:    opts.Images,
996
		Format:    json.RawMessage(opts.Format),
997
998
999
		System:    opts.System,
		Options:   opts.Options,
		KeepAlive: opts.KeepAlive,
Michael Yang's avatar
Michael Yang committed
1000
1001
1002
	}

	if err := client.Generate(ctx, &request, fn); err != nil {
1003
		if errors.Is(err, context.Canceled) {
1004
			return nil
1005
		}
1006
		return err
Patrick Devine's avatar
Patrick Devine committed
1007
	}
1008

1009
	if opts.Prompt != "" {
Michael Yang's avatar
Michael Yang committed
1010
1011
		fmt.Println()
		fmt.Println()
Patrick Devine's avatar
Patrick Devine committed
1012
	}
1013

1014
1015
1016
1017
	if !latest.Done {
		return nil
	}

Patrick Devine's avatar
Patrick Devine committed
1018
1019
	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
1020
		return err
Patrick Devine's avatar
Patrick Devine committed
1021
	}
Michael Yang's avatar
Michael Yang committed
1022

Patrick Devine's avatar
Patrick Devine committed
1023
1024
	if verbose {
		latest.Summary()
Michael Yang's avatar
Michael Yang committed
1025
	}
Michael Yang's avatar
Michael Yang committed
1026

Patrick Devine's avatar
Patrick Devine committed
1027
1028
1029
	ctx = context.WithValue(cmd.Context(), generateContextKey("context"), latest.Context)
	cmd.SetContext(ctx)

1030
	return nil
Michael Yang's avatar
Michael Yang committed
1031
1032
}

1033
func RunServer(_ *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1034
	if err := initializeKeypair(); err != nil {
1035
1036
1037
		return err
	}

Michael Yang's avatar
host  
Michael Yang committed
1038
	ln, err := net.Listen("tcp", envconfig.Host().Host)
1039
1040
1041
	if err != nil {
		return err
	}
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1042

1043
1044
1045
1046
1047
1048
	err = server.Serve(ln)
	if errors.Is(err, http.ErrServerClosed) {
		return nil
	}

	return err
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1049
1050
}

1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
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
1063
		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
1064
1065
1066
1067
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1068
		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
1069
1070
1071
1072
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1073
		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
1074
1075
1076
			return fmt.Errorf("could not create directory %w", err)
		}

Michael Yang's avatar
Michael Yang committed
1077
		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
1078
1079
1080
			return err
		}

Michael Yang's avatar
Michael Yang committed
1081
		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
1082
1083
1084
1085
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1086
		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
1087

Michael Yang's avatar
Michael Yang committed
1088
		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
1089
1090
1091
			return err
		}

Michael Yang's avatar
Michael Yang committed
1092
		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
1093
1094
1095
1096
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1097
func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1098
	client, err := api.ClientFromEnvironment()
1099
1100
1101
	if err != nil {
		return err
	}
Michael Yang's avatar
Michael Yang committed
1102
	if err := client.Heartbeat(cmd.Context()); err != nil {
1103
		if !strings.Contains(err.Error(), " refused") {
Bruce MacDonald's avatar
Bruce MacDonald committed
1104
1105
			return err
		}
1106
		if err := startApp(cmd.Context(), client); err != nil {
Michael Yang's avatar
lint  
Michael Yang committed
1107
			return errors.New("could not connect to ollama app, is it running?")
1108
1109
1110
1111
1112
		}
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1113
1114
1115
1116
1117
1118
1119
1120
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
1121
1122
1123
1124
1125
		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
1126
1127
	}

1128
	if serverVersion != version.Version {
Michael Yang's avatar
Michael Yang committed
1129
		fmt.Printf("Warning: client version is %s\n", version.Version)
1130
	}
Michael Yang's avatar
Michael Yang committed
1131
1132
}

1133
func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
1134
1135
1136
1137
1138
	if len(envs) == 0 {
		return
	}

	envUsage := `
1139
1140
Environment Variables:
`
1141
	for _, e := range envs {
1142
		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
1143
1144
1145
	}

	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
1146
1147
}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1148
1149
func NewCLI() *cobra.Command {
	log.SetFlags(log.LstdFlags | log.Lshortfile)
Michael Yang's avatar
Michael Yang committed
1150
	cobra.EnableCommandSorting = false
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1151

1152
	if runtime.GOOS == "windows" && term.IsTerminal(int(os.Stdout.Fd())) {
1153
		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
1154
1155
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1156
	rootCmd := &cobra.Command{
1157
1158
1159
1160
		Use:           "ollama",
		Short:         "Large language model runner",
		SilenceUsage:  true,
		SilenceErrors: true,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1161
1162
1163
		CompletionOptions: cobra.CompletionOptions{
			DisableDefaultCmd: true,
		},
Michael Yang's avatar
Michael Yang committed
1164
1165
1166
1167
1168
1169
1170
1171
		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
1172
1173
	}

Michael Yang's avatar
Michael Yang committed
1174
	rootCmd.Flags().BoolP("version", "v", false, "Show version information")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1175

1176
	createCmd := &cobra.Command{
1177
1178
		Use:     "create MODEL",
		Short:   "Create a model from a Modelfile",
Michael Yang's avatar
Michael Yang committed
1179
		Args:    cobra.ExactArgs(1),
1180
1181
		PreRunE: checkServerHeartbeat,
		RunE:    CreateHandler,
1182
1183
	}

1184
	createCmd.Flags().StringP("file", "f", "", "Name of the Modelfile (default \"Modelfile\"")
1185
	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_0)")
1186

Patrick Devine's avatar
Patrick Devine committed
1187
1188
1189
	showCmd := &cobra.Command{
		Use:     "show MODEL",
		Short:   "Show information for a model",
Michael Yang's avatar
Michael Yang committed
1190
		Args:    cobra.ExactArgs(1),
Patrick Devine's avatar
Patrick Devine committed
1191
1192
1193
1194
1195
1196
1197
1198
		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")
1199
	showCmd.Flags().Bool("system", false, "Show system message of a model")
Patrick Devine's avatar
Patrick Devine committed
1200

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1201
	runCmd := &cobra.Command{
1202
1203
1204
1205
1206
		Use:     "run MODEL [PROMPT]",
		Short:   "Run a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    RunHandler,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1207
1208
	}

1209
	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
1210
	runCmd.Flags().Bool("verbose", false, "Show timings for response")
1211
	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
1212
	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1213
	runCmd.Flags().String("format", "", "Response format (e.g. json)")
Patrick Devine's avatar
Patrick Devine committed
1214
1215
1216
1217
1218
1219
1220
1221
1222

	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
1223
1224
1225
1226
	serveCmd := &cobra.Command{
		Use:     "serve",
		Aliases: []string{"start"},
		Short:   "Start ollama",
Michael Yang's avatar
Michael Yang committed
1227
		Args:    cobra.ExactArgs(0),
Michael Yang's avatar
Michael Yang committed
1228
		RunE:    RunServer,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1229
1230
	}

1231
	pullCmd := &cobra.Command{
1232
1233
		Use:     "pull MODEL",
		Short:   "Pull a model from a registry",
Michael Yang's avatar
Michael Yang committed
1234
		Args:    cobra.ExactArgs(1),
1235
1236
		PreRunE: checkServerHeartbeat,
		RunE:    PullHandler,
1237
1238
	}

1239
1240
	pullCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1241
	pushCmd := &cobra.Command{
1242
1243
		Use:     "push MODEL",
		Short:   "Push a model to a registry",
Michael Yang's avatar
Michael Yang committed
1244
		Args:    cobra.ExactArgs(1),
1245
1246
		PreRunE: checkServerHeartbeat,
		RunE:    PushHandler,
1247
1248
	}

1249
1250
	pushCmd.Flags().Bool("insecure", false, "Use an insecure registry")

Patrick Devine's avatar
Patrick Devine committed
1251
	listCmd := &cobra.Command{
1252
		Use:     "list",
Patrick Devine's avatar
Patrick Devine committed
1253
		Aliases: []string{"ls"},
1254
		Short:   "List models",
1255
		PreRunE: checkServerHeartbeat,
1256
		RunE:    ListHandler,
1257
	}
1258
1259
1260
1261
1262
1263
1264
1265

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

Patrick Devine's avatar
Patrick Devine committed
1266
	copyCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1267
		Use:     "cp SOURCE DESTINATION",
1268
		Short:   "Copy a model",
Michael Yang's avatar
Michael Yang committed
1269
		Args:    cobra.ExactArgs(2),
1270
1271
		PreRunE: checkServerHeartbeat,
		RunE:    CopyHandler,
Patrick Devine's avatar
Patrick Devine committed
1272
1273
	}

1274
	deleteCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1275
		Use:     "rm MODEL [MODEL...]",
1276
1277
1278
1279
		Short:   "Remove a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    DeleteHandler,
Patrick Devine's avatar
Patrick Devine committed
1280
1281
	}

1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
	runnerCmd := &cobra.Command{
		Use:    "runner",
		Short:  llama.PrintSystemInfo(),
		Hidden: true,
		RunE: func(cmd *cobra.Command, args []string) error {
			return runner.Execute(os.Args[1:])
		},
		FParseErrWhitelist: cobra.FParseErrWhitelist{UnknownFlags: true},
	}
	runnerCmd.SetHelpFunc(func(cmd *cobra.Command, args []string) {
		_ = runner.Execute(args[1:])
	})

1295
1296
1297
	envVars := envconfig.AsMap()

	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
1298

1299
1300
1301
1302
	for _, cmd := range []*cobra.Command{
		createCmd,
		showCmd,
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1303
		stopCmd,
1304
1305
1306
		pullCmd,
		pushCmd,
		listCmd,
1307
		psCmd,
1308
1309
		copyCmd,
		deleteCmd,
1310
		serveCmd,
1311
	} {
1312
1313
		switch cmd {
		case runCmd:
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
			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"],
1326
				envVars["OLLAMA_SCHED_SPREAD"],
1327
				envVars["OLLAMA_TMPDIR"],
1328
				envVars["OLLAMA_FLASH_ATTENTION"],
1329
				envVars["OLLAMA_KV_CACHE_TYPE"],
1330
				envVars["OLLAMA_LLM_LIBRARY"],
1331
				envVars["OLLAMA_GPU_OVERHEAD"],
1332
				envVars["OLLAMA_LOAD_TIMEOUT"],
1333
			})
1334
1335
1336
		default:
			appendEnvDocs(cmd, envs)
		}
1337
1338
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1339
1340
	rootCmd.AddCommand(
		serveCmd,
1341
		createCmd,
Patrick Devine's avatar
Patrick Devine committed
1342
		showCmd,
1343
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1344
		stopCmd,
1345
1346
		pullCmd,
		pushCmd,
Patrick Devine's avatar
Patrick Devine committed
1347
		listCmd,
1348
		psCmd,
Patrick Devine's avatar
Patrick Devine committed
1349
		copyCmd,
1350
		deleteCmd,
1351
		runnerCmd,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1352
1353
1354
1355
	)

	return rootCmd
}