cmd.go 30.4 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
38
	"github.com/ollama/ollama/llama"
	"github.com/ollama/ollama/llama/runner"
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
var errModelfileNotFound = errors.New("specified Modelfile wasn't found")
47
48
49
50
51
52
53
54
55
56

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

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

	absName, err := filepath.Abs(filename)
57
	if err != nil {
58
		return "", err
59
60
	}

61
	_, err = os.Stat(absName)
62
	if err != nil {
63
		return fn, err
64
	}
65

66
67
68
69
	return absName, nil
}

func CreateHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
70
71
72
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

73
74
75
76
77
78
79
80
81
82
	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
83
		return err
84
85
86
87
88
89
90
91
	} else {
		f, err := os.Open(filename)
		if err != nil {
			return err
		}

		reader = f
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
92
93
	}

94
	modelfile, err := parser.ParseFile(reader)
Michael Yang's avatar
Michael Yang committed
95
96
97
98
	if err != nil {
		return err
	}

99
100
101
102
103
	status := "gathering model components"
	spinner := progress.NewSpinner(status)
	p.Add(status, spinner)

	req, err := modelfile.CreateRequest()
Michael Yang's avatar
Michael Yang committed
104
105
106
	if err != nil {
		return err
	}
107
	spinner.Stop()
Michael Yang's avatar
Michael Yang committed
108

109
110
111
112
113
	req.Name = args[0]
	quantize, _ := cmd.Flags().GetString("quantize")
	if quantize != "" {
		req.Quantize = quantize
	}
114

115
116
117
118
119
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

120
121
122
123
	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
124
125
				return err
			}
126
127
128
129
			fileMap[filepath.Base(f)] = digest
		}
		req.Files = fileMap
	}
Michael Yang's avatar
Michael Yang committed
130

131
132
133
134
	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
135
136
				return err
			}
137
			fileMap[filepath.Base(f)] = digest
Michael Yang's avatar
Michael Yang committed
138
		}
139
		req.Adapters = fileMap
Michael Yang's avatar
Michael Yang committed
140
	}
Michael Yang's avatar
Michael Yang committed
141

Michael Yang's avatar
Michael Yang committed
142
	bars := make(map[string]*progress.Bar)
143
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
144
145
146
		if resp.Digest != "" {
			bar, ok := bars[resp.Digest]
			if !ok {
147
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
148
149
150
151
152
153
154
155
156
157
158
159
160
				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)
		}

161
162
163
		return nil
	}

164
	if err := client.Create(cmd.Context(), req, fn); err != nil {
165
166
167
168
169
170
		return err
	}

	return nil
}

171
172
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
173
174
175
176
	if err != nil {
		return "", err
	}

177
	bin, err := os.Open(realPath)
178
179
180
181
182
	if err != nil {
		return "", err
	}
	defer bin.Close()

183
184
185
186
187
188
189
190
	// Get file info to retrieve the size
	fileInfo, err := bin.Stat()
	if err != nil {
		return "", err
	}
	fileSize := fileInfo.Size()

	var pw progressWriter
191
192
193
194
	status := fmt.Sprintf("copying file %s 0%%", digest)
	spinner := progress.NewSpinner(status)
	p.Add(status, spinner)
	defer spinner.Stop()
195
196
197
198
199
200
201
202
203
204

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

	go func() {
		ticker := time.NewTicker(60 * time.Millisecond)
		defer ticker.Stop()
		for {
			select {
			case <-ticker.C:
205
				spinner.SetMessage(fmt.Sprintf("copying file %s %d%%", digest, int(100*pw.n.Load()/fileSize)))
206
			case <-done:
207
				spinner.SetMessage(fmt.Sprintf("copying file %s 100%%", digest))
208
209
210
211
212
213
				return
			}
		}
	}()

	if err = client.CreateBlob(cmd.Context(), digest, io.TeeReader(bin, &pw)); err != nil {
214
215
216
217
218
		return "", err
	}
	return digest, nil
}

219
220
221
222
223
224
225
226
227
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
228
229
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
259
260
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
}

261
func RunHandler(cmd *cobra.Command, args []string) error {
262
263
264
	interactive := true

	opts := runOptions{
265
266
267
		Model:    args[0],
		WordWrap: os.Getenv("TERM") == "xterm-256color",
		Options:  map[string]interface{}{},
268
269
270
271
272
273
274
275
	}

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

276
277
278
279
280
281
282
283
284
285
286
287
	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}
	}

288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
	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
	}
304
305
306
307
	// Be quiet if we're redirecting to a pipe or file
	if !term.IsTerminal(int(os.Stdout.Fd())) {
		interactive = false
	}
308
309
310
311
312
313
314

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

315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
	// 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
337
338
	}

339
	opts.MultiModal = len(info.ProjectorInfo) != 0
340
341
342
	opts.ParentModel = info.Details.ParentModel

	if interactive {
Patrick Devine's avatar
Patrick Devine committed
343
		if err := loadOrUnloadModel(cmd, &opts); err != nil {
Michael Yang's avatar
Michael Yang committed
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
			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()
			}
		}

359
360
361
		return generateInteractive(cmd, opts)
	}
	return generate(cmd, opts)
Bruce MacDonald's avatar
Bruce MacDonald committed
362
363
}

364
func PushHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
365
	client, err := api.ClientFromEnvironment()
366
367
368
	if err != nil {
		return err
	}
369

370
371
372
373
374
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
375
376
377
378
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)
379
380
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
381

382
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
383
		if resp.Digest != "" {
384
385
386
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
387
388
389

			bar, ok := bars[resp.Digest]
			if !ok {
390
				bar = progress.NewBar(fmt.Sprintf("pushing %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
391
392
393
394
395
396
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
397
398
399
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
400
401
402
403
404
405

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

406
407
408
		return nil
	}

Michael Yang's avatar
Michael Yang committed
409
	request := api.PushRequest{Name: args[0], Insecure: insecure}
410
411

	n := model.ParseName(args[0])
Michael Yang's avatar
Michael Yang committed
412
	if err := client.Push(cmd.Context(), &request, fn); err != nil {
413
414
415
416
417
418
		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
419
420
421
		return err
	}

422
	p.Stop()
423
	spinner.Stop()
424
425
426
427
428
429
430
431

	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
432
	return nil
433
434
}

435
func ListHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
436
	client, err := api.ClientFromEnvironment()
437
438
439
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
440

Michael Yang's avatar
Michael Yang committed
441
	models, err := client.List(cmd.Context())
Patrick Devine's avatar
Patrick Devine committed
442
443
444
445
446
447
448
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
449
		if len(args) == 0 || strings.HasPrefix(strings.ToLower(m.Name), strings.ToLower(args[0])) {
450
			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
451
		}
Patrick Devine's avatar
Patrick Devine committed
452
453
454
	}

	table := tablewriter.NewWriter(os.Stdout)
Patrick Devine's avatar
Patrick Devine committed
455
	table.SetHeader([]string{"NAME", "ID", "SIZE", "MODIFIED"})
Patrick Devine's avatar
Patrick Devine committed
456
457
458
459
460
	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
461
	table.SetTablePadding("    ")
Patrick Devine's avatar
Patrick Devine committed
462
463
464
465
466
467
	table.AppendBulk(data)
	table.Render()

	return nil
}

468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
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
496
497
498
499
500
501
502
503
504

			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})
505
506
507
508
509
510
511
512
513
514
		}
	}

	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
515
	table.SetTablePadding("    ")
516
517
518
519
520
521
	table.AppendBulk(data)
	table.Render()

	return nil
}

522
func DeleteHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
523
	client, err := api.ClientFromEnvironment()
524
525
526
	if err != nil {
		return err
	}
527

528
529
530
531
532
533
534
535
536
537
538
	// 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)
		}
	}

539
540
	for _, name := range args {
		req := api.DeleteRequest{Name: name}
Michael Yang's avatar
Michael Yang committed
541
		if err := client.Delete(cmd.Context(), &req); err != nil {
542
543
544
			return err
		}
		fmt.Printf("deleted '%s'\n", name)
545
546
547
548
	}
	return nil
}

Patrick Devine's avatar
Patrick Devine committed
549
func ShowHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
550
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
	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 {
596
		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
597
598
	}

599
600
601
602
603
	req := api.ShowRequest{Name: args[0]}
	resp, err := client.Show(cmd.Context(), &req)
	if err != nil {
		return err
	}
604

605
	if flagsSet == 1 {
606
607
608
609
610
611
612
613
		switch showType {
		case "license":
			fmt.Println(resp.License)
		case "modelfile":
			fmt.Println(resp.Modelfile)
		case "parameters":
			fmt.Println(resp.Parameters)
		case "system":
614
			fmt.Print(resp.System)
615
		case "template":
616
			fmt.Print(resp.Template)
617
618
619
		}

		return nil
Patrick Devine's avatar
Patrick Devine committed
620
621
	}

Michael Yang's avatar
Michael Yang committed
622
	return showInfo(resp, os.Stdout)
623
624
}

Michael Yang's avatar
Michael Yang committed
625
626
627
628
629
630
631
632
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("    ")
633

Michael Yang's avatar
Michael Yang committed
634
635
636
		switch header {
		case "Template", "System", "License":
			table.SetColWidth(100)
637
638
		}

Michael Yang's avatar
Michael Yang committed
639
640
641
		table.AppendBulk(rows())
		table.Render()
		fmt.Fprintln(w)
Patrick Devine's avatar
Patrick Devine committed
642
643
	}

Michael Yang's avatar
Michael Yang committed
644
645
646
647
648
649
650
651
652
653
654
655
656
657
	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
	})
658

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

Michael Yang's avatar
Michael Yang committed
670
671
672
673
674
675
676
677
678
679
	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
		})
680
681
	}

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

Michael Yang's avatar
Michael Yang committed
692
693
694
695
696
	if resp.System != "" {
		tableRender("System", func() [][]string {
			return head(resp.System, 2)
		})
	}
697

Michael Yang's avatar
Michael Yang committed
698
699
700
701
	if resp.License != "" {
		tableRender("License", func() [][]string {
			return head(resp.License, 2)
		})
702
	}
Michael Yang's avatar
Michael Yang committed
703
704

	return nil
705
706
}

Patrick Devine's avatar
Patrick Devine committed
707
func CopyHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
708
	client, err := api.ClientFromEnvironment()
709
710
711
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
712
713

	req := api.CopyRequest{Source: args[0], Destination: args[1]}
Michael Yang's avatar
Michael Yang committed
714
	if err := client.Copy(cmd.Context(), &req); err != nil {
Patrick Devine's avatar
Patrick Devine committed
715
716
717
718
719
720
		return err
	}
	fmt.Printf("copied '%s' to '%s'\n", args[0], args[1])
	return nil
}

721
func PullHandler(cmd *cobra.Command, args []string) error {
722
723
724
725
726
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
727
	client, err := api.ClientFromEnvironment()
728
729
730
	if err != nil {
		return err
	}
731

Michael Yang's avatar
Michael Yang committed
732
733
734
735
736
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

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

737
738
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
739

740
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
741
		if resp.Digest != "" {
742
743
744
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
745
746
747

			bar, ok := bars[resp.Digest]
			if !ok {
748
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
749
750
751
752
753
754
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
755
756
757
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
758
759
760
761
762
763

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

764
765
		return nil
	}
766

Michael Yang's avatar
Michael Yang committed
767
	request := api.PullRequest{Name: args[0], Insecure: insecure}
Michael Yang's avatar
Michael Yang committed
768
	if err := client.Pull(cmd.Context(), &request, fn); err != nil {
Michael Yang's avatar
Michael Yang committed
769
770
771
772
		return err
	}

	return nil
Michael Yang's avatar
Michael Yang committed
773
774
}

775
776
type generateContextKey string

777
type runOptions struct {
778
779
780
781
782
783
784
785
786
787
	Model       string
	ParentModel string
	Prompt      string
	Messages    []api.Message
	WordWrap    bool
	Format      string
	System      string
	Images      []api.ImageData
	Options     map[string]interface{}
	MultiModal  bool
788
	KeepAlive   *api.Duration
789
790
}

791
792
793
794
795
796
797
798
799
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
800
801
			if state.lineLength+1 > termWidth-5 {
				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
802
803
804
805
806
807
808
					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
809
810
				a := runewidth.StringWidth(state.wordBuffer)
				if a > 0 {
811
					fmt.Printf("\x1b[%dD", a)
812
813
				}
				fmt.Printf("\x1b[K\n")
814
				fmt.Printf("%s%c", state.wordBuffer, ch)
815
816
817
				chWidth := runewidth.RuneWidth(ch)

				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
818
819
			} else {
				fmt.Print(string(ch))
820
821
822
823
				state.lineLength += runewidth.RuneWidth(ch)
				if runewidth.RuneWidth(ch) >= 2 {
					state.wordBuffer = ""
					continue
Josh Yan's avatar
Josh Yan committed
824
				}
825
826
827
828
829
830
831
832
833
834
835
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

				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
	}

886
887
888
889
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

890
891
892
	req := &api.ChatRequest{
		Model:    opts.Model,
		Messages: opts.Messages,
893
		Format:   json.RawMessage(opts.Format),
894
895
896
		Options:  opts.Options,
	}

897
898
899
900
	if opts.KeepAlive != nil {
		req.KeepAlive = opts.KeepAlive
	}

901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
	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
926
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
927
	if err != nil {
928
		return err
Patrick Devine's avatar
Patrick Devine committed
929
	}
Michael Yang's avatar
Michael Yang committed
930

Michael Yang's avatar
Michael Yang committed
931
	p := progress.NewProgress(os.Stderr)
932
	defer p.StopAndClear()
933

Michael Yang's avatar
Michael Yang committed
934
935
936
	spinner := progress.NewSpinner("")
	p.Add("", spinner)

937
938
939
940
941
942
943
	var latest api.GenerateResponse

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

Michael Yang's avatar
Michael Yang committed
944
	ctx, cancel := context.WithCancel(cmd.Context())
945
946
947
948
949
950
951
952
953
954
	defer cancel()

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

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

955
	var state *displayResponseState = &displayResponseState{}
956

957
	fn := func(response api.GenerateResponse) error {
Michael Yang's avatar
Michael Yang committed
958
		p.StopAndClear()
959

Patrick Devine's avatar
Patrick Devine committed
960
		latest = response
961
		content := response.Response
962

963
		displayResponse(content, opts.WordWrap, state)
964

Patrick Devine's avatar
Patrick Devine committed
965
966
		return nil
	}
967

968
969
970
971
972
973
974
	if opts.MultiModal {
		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
		if err != nil {
			return err
		}
	}

975
976
977
978
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

Michael Yang's avatar
Michael Yang committed
979
	request := api.GenerateRequest{
980
981
982
983
		Model:     opts.Model,
		Prompt:    opts.Prompt,
		Context:   generateContext,
		Images:    opts.Images,
984
		Format:    json.RawMessage(opts.Format),
985
986
987
		System:    opts.System,
		Options:   opts.Options,
		KeepAlive: opts.KeepAlive,
Michael Yang's avatar
Michael Yang committed
988
989
990
	}

	if err := client.Generate(ctx, &request, fn); err != nil {
991
		if errors.Is(err, context.Canceled) {
992
			return nil
993
		}
994
		return err
Patrick Devine's avatar
Patrick Devine committed
995
	}
996

997
	if opts.Prompt != "" {
Michael Yang's avatar
Michael Yang committed
998
999
		fmt.Println()
		fmt.Println()
Patrick Devine's avatar
Patrick Devine committed
1000
	}
1001

1002
1003
1004
1005
	if !latest.Done {
		return nil
	}

Patrick Devine's avatar
Patrick Devine committed
1006
1007
	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
1008
		return err
Patrick Devine's avatar
Patrick Devine committed
1009
	}
Michael Yang's avatar
Michael Yang committed
1010

Patrick Devine's avatar
Patrick Devine committed
1011
1012
	if verbose {
		latest.Summary()
Michael Yang's avatar
Michael Yang committed
1013
	}
Michael Yang's avatar
Michael Yang committed
1014

Patrick Devine's avatar
Patrick Devine committed
1015
1016
1017
	ctx = context.WithValue(cmd.Context(), generateContextKey("context"), latest.Context)
	cmd.SetContext(ctx)

1018
	return nil
Michael Yang's avatar
Michael Yang committed
1019
1020
}

1021
func RunServer(_ *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1022
	if err := initializeKeypair(); err != nil {
1023
1024
1025
		return err
	}

Michael Yang's avatar
host  
Michael Yang committed
1026
	ln, err := net.Listen("tcp", envconfig.Host().Host)
1027
1028
1029
	if err != nil {
		return err
	}
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1030

1031
1032
1033
1034
1035
1036
	err = server.Serve(ln)
	if errors.Is(err, http.ErrServerClosed) {
		return nil
	}

	return err
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1037
1038
}

1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
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
1051
		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
1052
1053
1054
1055
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1056
		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
1057
1058
1059
1060
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1061
		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
1062
1063
1064
			return fmt.Errorf("could not create directory %w", err)
		}

Michael Yang's avatar
Michael Yang committed
1065
		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
1066
1067
1068
			return err
		}

Michael Yang's avatar
Michael Yang committed
1069
		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
1070
1071
1072
1073
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1074
		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
1075

Michael Yang's avatar
Michael Yang committed
1076
		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
1077
1078
1079
			return err
		}

Michael Yang's avatar
Michael Yang committed
1080
		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
1081
1082
1083
1084
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1085
func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1086
	client, err := api.ClientFromEnvironment()
1087
1088
1089
	if err != nil {
		return err
	}
Michael Yang's avatar
Michael Yang committed
1090
	if err := client.Heartbeat(cmd.Context()); err != nil {
1091
		if !strings.Contains(err.Error(), " refused") {
Bruce MacDonald's avatar
Bruce MacDonald committed
1092
1093
			return err
		}
1094
		if err := startApp(cmd.Context(), client); err != nil {
Michael Yang's avatar
lint  
Michael Yang committed
1095
			return errors.New("could not connect to ollama app, is it running?")
1096
1097
1098
1099
1100
		}
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1101
1102
1103
1104
1105
1106
1107
1108
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
1109
1110
1111
1112
1113
		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
1114
1115
	}

1116
	if serverVersion != version.Version {
Michael Yang's avatar
Michael Yang committed
1117
		fmt.Printf("Warning: client version is %s\n", version.Version)
1118
	}
Michael Yang's avatar
Michael Yang committed
1119
1120
}

1121
func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
1122
1123
1124
1125
1126
	if len(envs) == 0 {
		return
	}

	envUsage := `
1127
1128
Environment Variables:
`
1129
	for _, e := range envs {
1130
		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
1131
1132
1133
	}

	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
1134
1135
}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1136
1137
func NewCLI() *cobra.Command {
	log.SetFlags(log.LstdFlags | log.Lshortfile)
Michael Yang's avatar
Michael Yang committed
1138
	cobra.EnableCommandSorting = false
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1139

1140
	if runtime.GOOS == "windows" && term.IsTerminal(int(os.Stdout.Fd())) {
1141
		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
1142
1143
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1144
	rootCmd := &cobra.Command{
1145
1146
1147
1148
		Use:           "ollama",
		Short:         "Large language model runner",
		SilenceUsage:  true,
		SilenceErrors: true,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1149
1150
1151
		CompletionOptions: cobra.CompletionOptions{
			DisableDefaultCmd: true,
		},
Michael Yang's avatar
Michael Yang committed
1152
1153
1154
1155
1156
1157
1158
1159
		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
1160
1161
	}

Michael Yang's avatar
Michael Yang committed
1162
	rootCmd.Flags().BoolP("version", "v", false, "Show version information")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1163

1164
	createCmd := &cobra.Command{
1165
1166
		Use:     "create MODEL",
		Short:   "Create a model from a Modelfile",
Michael Yang's avatar
Michael Yang committed
1167
		Args:    cobra.ExactArgs(1),
1168
1169
		PreRunE: checkServerHeartbeat,
		RunE:    CreateHandler,
1170
1171
	}

1172
	createCmd.Flags().StringP("file", "f", "", "Name of the Modelfile (default \"Modelfile\"")
1173
	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_0)")
1174

Patrick Devine's avatar
Patrick Devine committed
1175
1176
1177
	showCmd := &cobra.Command{
		Use:     "show MODEL",
		Short:   "Show information for a model",
Michael Yang's avatar
Michael Yang committed
1178
		Args:    cobra.ExactArgs(1),
Patrick Devine's avatar
Patrick Devine committed
1179
1180
1181
1182
1183
1184
1185
1186
		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")
1187
	showCmd.Flags().Bool("system", false, "Show system message of a model")
Patrick Devine's avatar
Patrick Devine committed
1188

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1189
	runCmd := &cobra.Command{
1190
1191
1192
1193
1194
		Use:     "run MODEL [PROMPT]",
		Short:   "Run a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    RunHandler,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1195
1196
	}

1197
	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
1198
	runCmd.Flags().Bool("verbose", false, "Show timings for response")
1199
	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
1200
	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1201
	runCmd.Flags().String("format", "", "Response format (e.g. json)")
Patrick Devine's avatar
Patrick Devine committed
1202
1203
1204
1205
1206
1207
1208
1209
1210

	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
1211
1212
1213
1214
	serveCmd := &cobra.Command{
		Use:     "serve",
		Aliases: []string{"start"},
		Short:   "Start ollama",
Michael Yang's avatar
Michael Yang committed
1215
		Args:    cobra.ExactArgs(0),
Michael Yang's avatar
Michael Yang committed
1216
		RunE:    RunServer,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1217
1218
	}

1219
	pullCmd := &cobra.Command{
1220
1221
		Use:     "pull MODEL",
		Short:   "Pull a model from a registry",
Michael Yang's avatar
Michael Yang committed
1222
		Args:    cobra.ExactArgs(1),
1223
1224
		PreRunE: checkServerHeartbeat,
		RunE:    PullHandler,
1225
1226
	}

1227
1228
	pullCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1229
	pushCmd := &cobra.Command{
1230
1231
		Use:     "push MODEL",
		Short:   "Push a model to a registry",
Michael Yang's avatar
Michael Yang committed
1232
		Args:    cobra.ExactArgs(1),
1233
1234
		PreRunE: checkServerHeartbeat,
		RunE:    PushHandler,
1235
1236
	}

1237
1238
	pushCmd.Flags().Bool("insecure", false, "Use an insecure registry")

Patrick Devine's avatar
Patrick Devine committed
1239
	listCmd := &cobra.Command{
1240
		Use:     "list",
Patrick Devine's avatar
Patrick Devine committed
1241
		Aliases: []string{"ls"},
1242
		Short:   "List models",
1243
		PreRunE: checkServerHeartbeat,
1244
		RunE:    ListHandler,
1245
	}
1246
1247
1248
1249
1250
1251
1252
1253

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

Patrick Devine's avatar
Patrick Devine committed
1254
	copyCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1255
		Use:     "cp SOURCE DESTINATION",
1256
		Short:   "Copy a model",
Michael Yang's avatar
Michael Yang committed
1257
		Args:    cobra.ExactArgs(2),
1258
1259
		PreRunE: checkServerHeartbeat,
		RunE:    CopyHandler,
Patrick Devine's avatar
Patrick Devine committed
1260
1261
	}

1262
	deleteCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1263
		Use:     "rm MODEL [MODEL...]",
1264
1265
1266
1267
		Short:   "Remove a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    DeleteHandler,
Patrick Devine's avatar
Patrick Devine committed
1268
1269
	}

1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
	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:])
	})

1283
1284
1285
	envVars := envconfig.AsMap()

	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
1286

1287
1288
1289
1290
	for _, cmd := range []*cobra.Command{
		createCmd,
		showCmd,
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1291
		stopCmd,
1292
1293
1294
		pullCmd,
		pushCmd,
		listCmd,
1295
		psCmd,
1296
1297
		copyCmd,
		deleteCmd,
1298
		serveCmd,
1299
	} {
1300
1301
		switch cmd {
		case runCmd:
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
			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"],
1314
				envVars["OLLAMA_SCHED_SPREAD"],
1315
				envVars["OLLAMA_TMPDIR"],
1316
				envVars["OLLAMA_FLASH_ATTENTION"],
1317
				envVars["OLLAMA_KV_CACHE_TYPE"],
1318
				envVars["OLLAMA_LLM_LIBRARY"],
1319
				envVars["OLLAMA_GPU_OVERHEAD"],
1320
				envVars["OLLAMA_LOAD_TIMEOUT"],
1321
			})
1322
1323
1324
		default:
			appendEnvDocs(cmd, envs)
		}
1325
1326
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1327
1328
	rootCmd.AddCommand(
		serveCmd,
1329
		createCmd,
Patrick Devine's avatar
Patrick Devine committed
1330
		showCmd,
1331
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1332
		stopCmd,
1333
1334
		pullCmd,
		pushCmd,
Patrick Devine's avatar
Patrick Devine committed
1335
		listCmd,
1336
		psCmd,
Patrick Devine's avatar
Patrick Devine committed
1337
		copyCmd,
1338
		deleteCmd,
1339
		runnerCmd,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1340
1341
1342
1343
	)

	return rootCmd
}