cmd.go 30.6 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

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 filename, 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
259
260
261
262
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
}

263
func RunHandler(cmd *cobra.Command, args []string) error {
264
265
266
	interactive := true

	opts := runOptions{
267
268
269
		Model:    args[0],
		WordWrap: os.Getenv("TERM") == "xterm-256color",
		Options:  map[string]interface{}{},
270
271
272
273
274
275
276
277
	}

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

278
279
280
281
282
283
284
285
286
287
288
289
	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}
	}

290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
	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
	}
306
307
308
309
	// Be quiet if we're redirecting to a pipe or file
	if !term.IsTerminal(int(os.Stdout.Fd())) {
		interactive = false
	}
310
311
312
313
314
315
316

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

317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
	// 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
339
340
	}

341
	opts.MultiModal = len(info.ProjectorInfo) != 0
342
343
344
	opts.ParentModel = info.Details.ParentModel

	if interactive {
Patrick Devine's avatar
Patrick Devine committed
345
		if err := loadOrUnloadModel(cmd, &opts); err != nil {
Michael Yang's avatar
Michael Yang committed
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
			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()
			}
		}

361
362
363
		return generateInteractive(cmd, opts)
	}
	return generate(cmd, opts)
Bruce MacDonald's avatar
Bruce MacDonald committed
364
365
}

366
func PushHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
367
	client, err := api.ClientFromEnvironment()
368
369
370
	if err != nil {
		return err
	}
371

372
373
374
375
376
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
377
378
379
380
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)
381
382
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
383

384
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
385
		if resp.Digest != "" {
386
387
388
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
389
390
391

			bar, ok := bars[resp.Digest]
			if !ok {
392
				bar = progress.NewBar(fmt.Sprintf("pushing %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
393
394
395
396
397
398
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
399
400
401
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
402
403
404
405
406
407

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

408
409
410
		return nil
	}

Michael Yang's avatar
Michael Yang committed
411
	request := api.PushRequest{Name: args[0], Insecure: insecure}
412
413

	n := model.ParseName(args[0])
Michael Yang's avatar
Michael Yang committed
414
	if err := client.Push(cmd.Context(), &request, fn); err != nil {
415
416
417
418
419
420
		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
421
422
423
		return err
	}

424
	p.Stop()
425
	spinner.Stop()
426
427
428
429
430
431
432
433

	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
434
	return nil
435
436
}

437
func ListHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
438
	client, err := api.ClientFromEnvironment()
439
440
441
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
442

Michael Yang's avatar
Michael Yang committed
443
	models, err := client.List(cmd.Context())
Patrick Devine's avatar
Patrick Devine committed
444
445
446
447
448
449
450
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
451
		if len(args) == 0 || strings.HasPrefix(strings.ToLower(m.Name), strings.ToLower(args[0])) {
452
			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
453
		}
Patrick Devine's avatar
Patrick Devine committed
454
455
456
	}

	table := tablewriter.NewWriter(os.Stdout)
Patrick Devine's avatar
Patrick Devine committed
457
	table.SetHeader([]string{"NAME", "ID", "SIZE", "MODIFIED"})
Patrick Devine's avatar
Patrick Devine committed
458
459
460
461
462
	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
463
	table.SetTablePadding("    ")
Patrick Devine's avatar
Patrick Devine committed
464
465
466
467
468
469
	table.AppendBulk(data)
	table.Render()

	return nil
}

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
496
497
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
498
499
500
501
502
503
504
505
506

			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})
507
508
509
510
511
512
513
514
515
516
		}
	}

	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
517
	table.SetTablePadding("    ")
518
519
520
521
522
523
	table.AppendBulk(data)
	table.Render()

	return nil
}

524
func DeleteHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
525
	client, err := api.ClientFromEnvironment()
526
527
528
	if err != nil {
		return err
	}
529

530
531
532
533
534
535
536
537
538
539
540
	// 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)
		}
	}

541
542
	for _, name := range args {
		req := api.DeleteRequest{Name: name}
Michael Yang's avatar
Michael Yang committed
543
		if err := client.Delete(cmd.Context(), &req); err != nil {
544
545
546
			return err
		}
		fmt.Printf("deleted '%s'\n", name)
547
548
549
550
	}
	return nil
}

Patrick Devine's avatar
Patrick Devine committed
551
func ShowHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
552
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
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
596
597
	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 {
598
		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
599
600
	}

601
602
603
604
605
	req := api.ShowRequest{Name: args[0]}
	resp, err := client.Show(cmd.Context(), &req)
	if err != nil {
		return err
	}
606

607
	if flagsSet == 1 {
608
609
610
611
612
613
614
615
		switch showType {
		case "license":
			fmt.Println(resp.License)
		case "modelfile":
			fmt.Println(resp.Modelfile)
		case "parameters":
			fmt.Println(resp.Parameters)
		case "system":
616
			fmt.Print(resp.System)
617
		case "template":
618
			fmt.Print(resp.Template)
619
620
621
		}

		return nil
Patrick Devine's avatar
Patrick Devine committed
622
623
	}

Michael Yang's avatar
Michael Yang committed
624
	return showInfo(resp, os.Stdout)
625
626
}

Michael Yang's avatar
Michael Yang committed
627
628
629
630
631
632
633
634
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("    ")
635

Michael Yang's avatar
Michael Yang committed
636
637
638
		switch header {
		case "Template", "System", "License":
			table.SetColWidth(100)
639
640
		}

Michael Yang's avatar
Michael Yang committed
641
642
643
		table.AppendBulk(rows())
		table.Render()
		fmt.Fprintln(w)
Patrick Devine's avatar
Patrick Devine committed
644
645
	}

Michael Yang's avatar
Michael Yang committed
646
647
648
649
650
651
652
653
654
655
656
657
658
659
	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
	})
660

Michael Yang's avatar
Michael Yang committed
661
662
663
664
665
666
667
668
669
	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
		})
670
671
	}

Michael Yang's avatar
Michael Yang committed
672
673
674
675
676
677
678
679
680
681
	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
		})
682
683
	}

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

Michael Yang's avatar
Michael Yang committed
694
695
696
697
698
	if resp.System != "" {
		tableRender("System", func() [][]string {
			return head(resp.System, 2)
		})
	}
699

Michael Yang's avatar
Michael Yang committed
700
701
702
703
	if resp.License != "" {
		tableRender("License", func() [][]string {
			return head(resp.License, 2)
		})
704
	}
Michael Yang's avatar
Michael Yang committed
705
706

	return nil
707
708
}

Patrick Devine's avatar
Patrick Devine committed
709
func CopyHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
710
	client, err := api.ClientFromEnvironment()
711
712
713
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
714
715

	req := api.CopyRequest{Source: args[0], Destination: args[1]}
Michael Yang's avatar
Michael Yang committed
716
	if err := client.Copy(cmd.Context(), &req); err != nil {
Patrick Devine's avatar
Patrick Devine committed
717
718
719
720
721
722
		return err
	}
	fmt.Printf("copied '%s' to '%s'\n", args[0], args[1])
	return nil
}

723
func PullHandler(cmd *cobra.Command, args []string) error {
724
725
726
727
728
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
729
	client, err := api.ClientFromEnvironment()
730
731
732
	if err != nil {
		return err
	}
733

Michael Yang's avatar
Michael Yang committed
734
735
736
737
738
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

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

739
740
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
741

742
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
743
		if resp.Digest != "" {
744
745
746
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
747
748
749

			bar, ok := bars[resp.Digest]
			if !ok {
750
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
751
752
753
754
755
756
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
757
758
759
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
760
761
762
763
764
765

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

766
767
		return nil
	}
768

Michael Yang's avatar
Michael Yang committed
769
	request := api.PullRequest{Name: args[0], Insecure: insecure}
Michael Yang's avatar
Michael Yang committed
770
	if err := client.Pull(cmd.Context(), &request, fn); err != nil {
Michael Yang's avatar
Michael Yang committed
771
772
773
774
		return err
	}

	return nil
Michael Yang's avatar
Michael Yang committed
775
776
}

777
778
type generateContextKey string

779
type runOptions struct {
780
781
782
783
784
785
786
787
788
789
	Model       string
	ParentModel string
	Prompt      string
	Messages    []api.Message
	WordWrap    bool
	Format      string
	System      string
	Images      []api.ImageData
	Options     map[string]interface{}
	MultiModal  bool
790
	KeepAlive   *api.Duration
791
792
}

793
794
795
796
797
798
799
800
801
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
802
803
			if state.lineLength+1 > termWidth-5 {
				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
804
805
806
807
808
809
810
					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
811
812
				a := runewidth.StringWidth(state.wordBuffer)
				if a > 0 {
813
					fmt.Printf("\x1b[%dD", a)
814
815
				}
				fmt.Printf("\x1b[K\n")
816
				fmt.Printf("%s%c", state.wordBuffer, ch)
817
818
819
				chWidth := runewidth.RuneWidth(ch)

				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
820
821
			} else {
				fmt.Print(string(ch))
822
823
824
825
				state.lineLength += runewidth.RuneWidth(ch)
				if runewidth.RuneWidth(ch) >= 2 {
					state.wordBuffer = ""
					continue
Josh Yan's avatar
Josh Yan committed
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
886
887

				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
	}

888
889
890
891
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

892
893
894
	req := &api.ChatRequest{
		Model:    opts.Model,
		Messages: opts.Messages,
895
		Format:   json.RawMessage(opts.Format),
896
897
898
		Options:  opts.Options,
	}

899
900
901
902
	if opts.KeepAlive != nil {
		req.KeepAlive = opts.KeepAlive
	}

903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
	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
928
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
929
	if err != nil {
930
		return err
Patrick Devine's avatar
Patrick Devine committed
931
	}
Michael Yang's avatar
Michael Yang committed
932

Michael Yang's avatar
Michael Yang committed
933
	p := progress.NewProgress(os.Stderr)
934
	defer p.StopAndClear()
935

Michael Yang's avatar
Michael Yang committed
936
937
938
	spinner := progress.NewSpinner("")
	p.Add("", spinner)

939
940
941
942
943
944
945
	var latest api.GenerateResponse

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

Michael Yang's avatar
Michael Yang committed
946
	ctx, cancel := context.WithCancel(cmd.Context())
947
948
949
950
951
952
953
954
955
956
	defer cancel()

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

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

957
	var state *displayResponseState = &displayResponseState{}
958

959
	fn := func(response api.GenerateResponse) error {
Michael Yang's avatar
Michael Yang committed
960
		p.StopAndClear()
961

Patrick Devine's avatar
Patrick Devine committed
962
		latest = response
963
		content := response.Response
964

965
		displayResponse(content, opts.WordWrap, state)
966

Patrick Devine's avatar
Patrick Devine committed
967
968
		return nil
	}
969

970
971
972
973
974
975
976
	if opts.MultiModal {
		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
		if err != nil {
			return err
		}
	}

977
978
979
980
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

Michael Yang's avatar
Michael Yang committed
981
	request := api.GenerateRequest{
982
983
984
985
		Model:     opts.Model,
		Prompt:    opts.Prompt,
		Context:   generateContext,
		Images:    opts.Images,
986
		Format:    json.RawMessage(opts.Format),
987
988
989
		System:    opts.System,
		Options:   opts.Options,
		KeepAlive: opts.KeepAlive,
Michael Yang's avatar
Michael Yang committed
990
991
992
	}

	if err := client.Generate(ctx, &request, fn); err != nil {
993
		if errors.Is(err, context.Canceled) {
994
			return nil
995
		}
996
		return err
Patrick Devine's avatar
Patrick Devine committed
997
	}
998

999
	if opts.Prompt != "" {
Michael Yang's avatar
Michael Yang committed
1000
1001
		fmt.Println()
		fmt.Println()
Patrick Devine's avatar
Patrick Devine committed
1002
	}
1003

1004
1005
1006
1007
	if !latest.Done {
		return nil
	}

Patrick Devine's avatar
Patrick Devine committed
1008
1009
	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
1010
		return err
Patrick Devine's avatar
Patrick Devine committed
1011
	}
Michael Yang's avatar
Michael Yang committed
1012

Patrick Devine's avatar
Patrick Devine committed
1013
1014
	if verbose {
		latest.Summary()
Michael Yang's avatar
Michael Yang committed
1015
	}
Michael Yang's avatar
Michael Yang committed
1016

Patrick Devine's avatar
Patrick Devine committed
1017
1018
1019
	ctx = context.WithValue(cmd.Context(), generateContextKey("context"), latest.Context)
	cmd.SetContext(ctx)

1020
	return nil
Michael Yang's avatar
Michael Yang committed
1021
1022
}

1023
func RunServer(_ *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1024
	if err := initializeKeypair(); err != nil {
1025
1026
1027
		return err
	}

Michael Yang's avatar
host  
Michael Yang committed
1028
	ln, err := net.Listen("tcp", envconfig.Host().Host)
1029
1030
1031
	if err != nil {
		return err
	}
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1032

1033
1034
1035
1036
1037
1038
	err = server.Serve(ln)
	if errors.Is(err, http.ErrServerClosed) {
		return nil
	}

	return err
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1039
1040
}

1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
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
1053
		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
1054
1055
1056
1057
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1058
		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
1059
1060
1061
1062
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1063
		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
1064
1065
1066
			return fmt.Errorf("could not create directory %w", err)
		}

Michael Yang's avatar
Michael Yang committed
1067
		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
1068
1069
1070
			return err
		}

Michael Yang's avatar
Michael Yang committed
1071
		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
1072
1073
1074
1075
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1076
		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
1077

Michael Yang's avatar
Michael Yang committed
1078
		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
1079
1080
1081
			return err
		}

Michael Yang's avatar
Michael Yang committed
1082
		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
1083
1084
1085
1086
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1087
func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1088
	client, err := api.ClientFromEnvironment()
1089
1090
1091
	if err != nil {
		return err
	}
Michael Yang's avatar
Michael Yang committed
1092
	if err := client.Heartbeat(cmd.Context()); err != nil {
1093
		if !strings.Contains(err.Error(), " refused") {
Bruce MacDonald's avatar
Bruce MacDonald committed
1094
1095
			return err
		}
1096
		if err := startApp(cmd.Context(), client); err != nil {
Michael Yang's avatar
lint  
Michael Yang committed
1097
			return errors.New("could not connect to ollama app, is it running?")
1098
1099
1100
1101
1102
		}
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1103
1104
1105
1106
1107
1108
1109
1110
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
1111
1112
1113
1114
1115
		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
1116
1117
	}

1118
	if serverVersion != version.Version {
Michael Yang's avatar
Michael Yang committed
1119
		fmt.Printf("Warning: client version is %s\n", version.Version)
1120
	}
Michael Yang's avatar
Michael Yang committed
1121
1122
}

1123
func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
1124
1125
1126
1127
1128
	if len(envs) == 0 {
		return
	}

	envUsage := `
1129
1130
Environment Variables:
`
1131
	for _, e := range envs {
1132
		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
1133
1134
1135
	}

	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
1136
1137
}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1138
1139
func NewCLI() *cobra.Command {
	log.SetFlags(log.LstdFlags | log.Lshortfile)
Michael Yang's avatar
Michael Yang committed
1140
	cobra.EnableCommandSorting = false
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1141

1142
	if runtime.GOOS == "windows" && term.IsTerminal(int(os.Stdout.Fd())) {
1143
		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
1144
1145
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1146
	rootCmd := &cobra.Command{
1147
1148
1149
1150
		Use:           "ollama",
		Short:         "Large language model runner",
		SilenceUsage:  true,
		SilenceErrors: true,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1151
1152
1153
		CompletionOptions: cobra.CompletionOptions{
			DisableDefaultCmd: true,
		},
Michael Yang's avatar
Michael Yang committed
1154
1155
1156
1157
1158
1159
1160
1161
		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
1162
1163
	}

Michael Yang's avatar
Michael Yang committed
1164
	rootCmd.Flags().BoolP("version", "v", false, "Show version information")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1165

1166
	createCmd := &cobra.Command{
1167
1168
		Use:     "create MODEL",
		Short:   "Create a model from a Modelfile",
Michael Yang's avatar
Michael Yang committed
1169
		Args:    cobra.ExactArgs(1),
1170
1171
		PreRunE: checkServerHeartbeat,
		RunE:    CreateHandler,
1172
1173
	}

1174
	createCmd.Flags().StringP("file", "f", "", "Name of the Modelfile (default \"Modelfile\"")
1175
	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_0)")
1176

Patrick Devine's avatar
Patrick Devine committed
1177
1178
1179
	showCmd := &cobra.Command{
		Use:     "show MODEL",
		Short:   "Show information for a model",
Michael Yang's avatar
Michael Yang committed
1180
		Args:    cobra.ExactArgs(1),
Patrick Devine's avatar
Patrick Devine committed
1181
1182
1183
1184
1185
1186
1187
1188
		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")
1189
	showCmd.Flags().Bool("system", false, "Show system message of a model")
Patrick Devine's avatar
Patrick Devine committed
1190

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1191
	runCmd := &cobra.Command{
1192
1193
1194
1195
1196
		Use:     "run MODEL [PROMPT]",
		Short:   "Run a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    RunHandler,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1197
1198
	}

1199
	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
1200
	runCmd.Flags().Bool("verbose", false, "Show timings for response")
1201
	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
1202
	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1203
	runCmd.Flags().String("format", "", "Response format (e.g. json)")
Patrick Devine's avatar
Patrick Devine committed
1204
1205
1206
1207
1208
1209
1210
1211
1212

	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
1213
1214
1215
1216
	serveCmd := &cobra.Command{
		Use:     "serve",
		Aliases: []string{"start"},
		Short:   "Start ollama",
Michael Yang's avatar
Michael Yang committed
1217
		Args:    cobra.ExactArgs(0),
Michael Yang's avatar
Michael Yang committed
1218
		RunE:    RunServer,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1219
1220
	}

1221
	pullCmd := &cobra.Command{
1222
1223
		Use:     "pull MODEL",
		Short:   "Pull a model from a registry",
Michael Yang's avatar
Michael Yang committed
1224
		Args:    cobra.ExactArgs(1),
1225
1226
		PreRunE: checkServerHeartbeat,
		RunE:    PullHandler,
1227
1228
	}

1229
1230
	pullCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1231
	pushCmd := &cobra.Command{
1232
1233
		Use:     "push MODEL",
		Short:   "Push a model to a registry",
Michael Yang's avatar
Michael Yang committed
1234
		Args:    cobra.ExactArgs(1),
1235
1236
		PreRunE: checkServerHeartbeat,
		RunE:    PushHandler,
1237
1238
	}

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

Patrick Devine's avatar
Patrick Devine committed
1241
	listCmd := &cobra.Command{
1242
		Use:     "list",
Patrick Devine's avatar
Patrick Devine committed
1243
		Aliases: []string{"ls"},
1244
		Short:   "List models",
1245
		PreRunE: checkServerHeartbeat,
1246
		RunE:    ListHandler,
1247
	}
1248
1249
1250
1251
1252
1253
1254
1255

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

Patrick Devine's avatar
Patrick Devine committed
1256
	copyCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1257
		Use:     "cp SOURCE DESTINATION",
1258
		Short:   "Copy a model",
Michael Yang's avatar
Michael Yang committed
1259
		Args:    cobra.ExactArgs(2),
1260
1261
		PreRunE: checkServerHeartbeat,
		RunE:    CopyHandler,
Patrick Devine's avatar
Patrick Devine committed
1262
1263
	}

1264
	deleteCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1265
		Use:     "rm MODEL [MODEL...]",
1266
1267
1268
1269
		Short:   "Remove a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    DeleteHandler,
Patrick Devine's avatar
Patrick Devine committed
1270
1271
	}

1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
	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:])
	})

1285
1286
1287
	envVars := envconfig.AsMap()

	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
1288

1289
1290
1291
1292
	for _, cmd := range []*cobra.Command{
		createCmd,
		showCmd,
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1293
		stopCmd,
1294
1295
1296
		pullCmd,
		pushCmd,
		listCmd,
1297
		psCmd,
1298
1299
		copyCmd,
		deleteCmd,
1300
		serveCmd,
1301
	} {
1302
1303
		switch cmd {
		case runCmd:
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
			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"],
1316
				envVars["OLLAMA_SCHED_SPREAD"],
1317
				envVars["OLLAMA_TMPDIR"],
1318
				envVars["OLLAMA_FLASH_ATTENTION"],
1319
				envVars["OLLAMA_KV_CACHE_TYPE"],
1320
				envVars["OLLAMA_LLM_LIBRARY"],
1321
				envVars["OLLAMA_GPU_OVERHEAD"],
1322
				envVars["OLLAMA_LOAD_TIMEOUT"],
1323
			})
1324
1325
1326
		default:
			appendEnvDocs(cmd, envs)
		}
1327
1328
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1329
1330
	rootCmd.AddCommand(
		serveCmd,
1331
		createCmd,
Patrick Devine's avatar
Patrick Devine committed
1332
		showCmd,
1333
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1334
		stopCmd,
1335
1336
		pullCmd,
		pushCmd,
Patrick Devine's avatar
Patrick Devine committed
1337
		listCmd,
1338
		psCmd,
Patrick Devine's avatar
Patrick Devine committed
1339
		copyCmd,
1340
		deleteCmd,
1341
		runnerCmd,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1342
1343
1344
1345
	)

	return rootCmd
}