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
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
		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")
		}
168
169
170
171
172
173
		return err
	}

	return nil
}

174
175
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
176
177
178
179
	if err != nil {
		return "", err
	}

180
	bin, err := os.Open(realPath)
181
182
183
184
185
	if err != nil {
		return "", err
	}
	defer bin.Close()

186
187
188
189
190
191
192
193
	// Get file info to retrieve the size
	fileInfo, err := bin.Stat()
	if err != nil {
		return "", err
	}
	fileSize := fileInfo.Size()

	var pw progressWriter
194
195
196
197
	status := fmt.Sprintf("copying file %s 0%%", digest)
	spinner := progress.NewSpinner(status)
	p.Add(status, spinner)
	defer spinner.Stop()
198
199
200
201
202
203
204
205
206
207

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

	go func() {
		ticker := time.NewTicker(60 * time.Millisecond)
		defer ticker.Stop()
		for {
			select {
			case <-ticker.C:
208
				spinner.SetMessage(fmt.Sprintf("copying file %s %d%%", digest, int(100*pw.n.Load()/fileSize)))
209
			case <-done:
210
				spinner.SetMessage(fmt.Sprintf("copying file %s 100%%", digest))
211
212
213
214
215
216
				return
			}
		}
	}()

	if err = client.CreateBlob(cmd.Context(), digest, io.TeeReader(bin, &pw)); err != nil {
217
218
219
220
221
		return "", err
	}
	return digest, nil
}

222
223
224
225
226
227
228
229
230
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
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
263
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
}

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
	opts.MultiModal = len(info.ProjectorInfo) != 0
343
344
345
	opts.ParentModel = info.Details.ParentModel

	if interactive {
Patrick Devine's avatar
Patrick Devine committed
346
		if err := loadOrUnloadModel(cmd, &opts); err != nil {
Michael Yang's avatar
Michael Yang committed
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
			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()
			}
		}

362
363
364
		return generateInteractive(cmd, opts)
	}
	return generate(cmd, opts)
Bruce MacDonald's avatar
Bruce MacDonald committed
365
366
}

367
func PushHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
368
	client, err := api.ClientFromEnvironment()
369
370
371
	if err != nil {
		return err
	}
372

373
374
375
376
377
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
378
379
380
381
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)
382
383
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
384

385
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
386
		if resp.Digest != "" {
387
388
389
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
390
391
392

			bar, ok := bars[resp.Digest]
			if !ok {
393
				bar = progress.NewBar(fmt.Sprintf("pushing %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
394
395
396
397
398
399
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
400
401
402
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
403
404
405
406
407
408

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

409
410
411
		return nil
	}

Michael Yang's avatar
Michael Yang committed
412
	request := api.PushRequest{Name: args[0], Insecure: insecure}
413
414

	n := model.ParseName(args[0])
Michael Yang's avatar
Michael Yang committed
415
	if err := client.Push(cmd.Context(), &request, fn); err != nil {
416
417
418
419
420
421
		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
422
423
424
		return err
	}

425
	p.Stop()
426
	spinner.Stop()
427
428
429
430
431
432
433
434

	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
435
	return nil
436
437
}

438
func ListHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
439
	client, err := api.ClientFromEnvironment()
440
441
442
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
443

Michael Yang's avatar
Michael Yang committed
444
	models, err := client.List(cmd.Context())
Patrick Devine's avatar
Patrick Devine committed
445
446
447
448
449
450
451
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
452
		if len(args) == 0 || strings.HasPrefix(strings.ToLower(m.Name), strings.ToLower(args[0])) {
453
			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
454
		}
Patrick Devine's avatar
Patrick Devine committed
455
456
457
	}

	table := tablewriter.NewWriter(os.Stdout)
Patrick Devine's avatar
Patrick Devine committed
458
	table.SetHeader([]string{"NAME", "ID", "SIZE", "MODIFIED"})
Patrick Devine's avatar
Patrick Devine committed
459
460
461
462
463
	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
464
	table.SetTablePadding("    ")
Patrick Devine's avatar
Patrick Devine committed
465
466
467
468
469
470
	table.AppendBulk(data)
	table.Render()

	return nil
}

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
498
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
499
500
501
502
503
504
505
506
507

			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})
508
509
510
511
512
513
514
515
516
517
		}
	}

	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
518
	table.SetTablePadding("    ")
519
520
521
522
523
524
	table.AppendBulk(data)
	table.Render()

	return nil
}

525
func DeleteHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
526
	client, err := api.ClientFromEnvironment()
527
528
529
	if err != nil {
		return err
	}
530

531
532
533
534
535
536
537
538
539
540
541
	// 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)
		}
	}

542
543
	for _, name := range args {
		req := api.DeleteRequest{Name: name}
Michael Yang's avatar
Michael Yang committed
544
		if err := client.Delete(cmd.Context(), &req); err != nil {
545
546
547
			return err
		}
		fmt.Printf("deleted '%s'\n", name)
548
549
550
551
	}
	return nil
}

Patrick Devine's avatar
Patrick Devine committed
552
func ShowHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
553
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
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
598
	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 {
599
		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
600
601
	}

602
603
604
605
606
	req := api.ShowRequest{Name: args[0]}
	resp, err := client.Show(cmd.Context(), &req)
	if err != nil {
		return err
	}
607

608
	if flagsSet == 1 {
609
610
611
612
613
614
615
616
		switch showType {
		case "license":
			fmt.Println(resp.License)
		case "modelfile":
			fmt.Println(resp.Modelfile)
		case "parameters":
			fmt.Println(resp.Parameters)
		case "system":
617
			fmt.Print(resp.System)
618
		case "template":
619
			fmt.Print(resp.Template)
620
621
622
		}

		return nil
Patrick Devine's avatar
Patrick Devine committed
623
624
	}

Michael Yang's avatar
Michael Yang committed
625
	return showInfo(resp, os.Stdout)
626
627
}

Michael Yang's avatar
Michael Yang committed
628
629
630
631
632
633
634
635
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("    ")
636

Michael Yang's avatar
Michael Yang committed
637
638
639
		switch header {
		case "Template", "System", "License":
			table.SetColWidth(100)
640
641
		}

Michael Yang's avatar
Michael Yang committed
642
643
644
		table.AppendBulk(rows())
		table.Render()
		fmt.Fprintln(w)
Patrick Devine's avatar
Patrick Devine committed
645
646
	}

Michael Yang's avatar
Michael Yang committed
647
648
649
650
651
652
653
654
655
656
657
658
659
660
	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
	})
661

Michael Yang's avatar
Michael Yang committed
662
663
664
665
666
667
668
669
670
	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
		})
671
672
	}

Michael Yang's avatar
Michael Yang committed
673
674
675
676
677
678
679
680
681
682
	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
		})
683
684
	}

Michael Yang's avatar
Michael Yang committed
685
686
687
688
689
	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)})
690
691
			}
		}
Michael Yang's avatar
Michael Yang committed
692
		return
693
694
	}

Michael Yang's avatar
Michael Yang committed
695
696
697
698
699
	if resp.System != "" {
		tableRender("System", func() [][]string {
			return head(resp.System, 2)
		})
	}
700

Michael Yang's avatar
Michael Yang committed
701
702
703
704
	if resp.License != "" {
		tableRender("License", func() [][]string {
			return head(resp.License, 2)
		})
705
	}
Michael Yang's avatar
Michael Yang committed
706
707

	return nil
708
709
}

Patrick Devine's avatar
Patrick Devine committed
710
func CopyHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
711
	client, err := api.ClientFromEnvironment()
712
713
714
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
715
716

	req := api.CopyRequest{Source: args[0], Destination: args[1]}
Michael Yang's avatar
Michael Yang committed
717
	if err := client.Copy(cmd.Context(), &req); err != nil {
Patrick Devine's avatar
Patrick Devine committed
718
719
720
721
722
723
		return err
	}
	fmt.Printf("copied '%s' to '%s'\n", args[0], args[1])
	return nil
}

724
func PullHandler(cmd *cobra.Command, args []string) error {
725
726
727
728
729
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
730
	client, err := api.ClientFromEnvironment()
731
732
733
	if err != nil {
		return err
	}
734

Michael Yang's avatar
Michael Yang committed
735
736
737
738
739
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

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

740
741
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
742

743
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
744
		if resp.Digest != "" {
745
746
747
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
748
749
750

			bar, ok := bars[resp.Digest]
			if !ok {
751
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
752
753
754
755
756
757
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
758
759
760
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
761
762
763
764
765
766

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

767
768
		return nil
	}
769

Michael Yang's avatar
Michael Yang committed
770
	request := api.PullRequest{Name: args[0], Insecure: insecure}
Michael Yang's avatar
Michael Yang committed
771
	if err := client.Pull(cmd.Context(), &request, fn); err != nil {
Michael Yang's avatar
Michael Yang committed
772
773
774
775
		return err
	}

	return nil
Michael Yang's avatar
Michael Yang committed
776
777
}

778
779
type generateContextKey string

780
type runOptions struct {
781
782
783
784
785
786
787
788
789
790
	Model       string
	ParentModel string
	Prompt      string
	Messages    []api.Message
	WordWrap    bool
	Format      string
	System      string
	Images      []api.ImageData
	Options     map[string]interface{}
	MultiModal  bool
791
	KeepAlive   *api.Duration
792
793
}

794
795
796
797
798
799
800
801
802
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
803
804
			if state.lineLength+1 > termWidth-5 {
				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
805
806
807
808
809
810
811
					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
812
813
				a := runewidth.StringWidth(state.wordBuffer)
				if a > 0 {
814
					fmt.Printf("\x1b[%dD", a)
815
816
				}
				fmt.Printf("\x1b[K\n")
817
				fmt.Printf("%s%c", state.wordBuffer, ch)
818
819
820
				chWidth := runewidth.RuneWidth(ch)

				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
821
822
			} else {
				fmt.Print(string(ch))
823
824
825
826
				state.lineLength += runewidth.RuneWidth(ch)
				if runewidth.RuneWidth(ch) >= 2 {
					state.wordBuffer = ""
					continue
Josh Yan's avatar
Josh Yan committed
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
888

				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
	}

889
890
891
892
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

893
894
895
	req := &api.ChatRequest{
		Model:    opts.Model,
		Messages: opts.Messages,
896
		Format:   json.RawMessage(opts.Format),
897
898
899
		Options:  opts.Options,
	}

900
901
902
903
	if opts.KeepAlive != nil {
		req.KeepAlive = opts.KeepAlive
	}

904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
	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
929
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
930
	if err != nil {
931
		return err
Patrick Devine's avatar
Patrick Devine committed
932
	}
Michael Yang's avatar
Michael Yang committed
933

Michael Yang's avatar
Michael Yang committed
934
	p := progress.NewProgress(os.Stderr)
935
	defer p.StopAndClear()
936

Michael Yang's avatar
Michael Yang committed
937
938
939
	spinner := progress.NewSpinner("")
	p.Add("", spinner)

940
941
942
943
944
945
946
	var latest api.GenerateResponse

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

Michael Yang's avatar
Michael Yang committed
947
	ctx, cancel := context.WithCancel(cmd.Context())
948
949
950
951
952
953
954
955
956
957
	defer cancel()

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

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

958
	var state *displayResponseState = &displayResponseState{}
959

960
	fn := func(response api.GenerateResponse) error {
Michael Yang's avatar
Michael Yang committed
961
		p.StopAndClear()
962

Patrick Devine's avatar
Patrick Devine committed
963
		latest = response
964
		content := response.Response
965

966
		displayResponse(content, opts.WordWrap, state)
967

Patrick Devine's avatar
Patrick Devine committed
968
969
		return nil
	}
970

971
972
973
974
975
976
977
	if opts.MultiModal {
		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
		if err != nil {
			return err
		}
	}

978
979
980
981
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

Michael Yang's avatar
Michael Yang committed
982
	request := api.GenerateRequest{
983
984
985
986
		Model:     opts.Model,
		Prompt:    opts.Prompt,
		Context:   generateContext,
		Images:    opts.Images,
987
		Format:    json.RawMessage(opts.Format),
988
989
990
		System:    opts.System,
		Options:   opts.Options,
		KeepAlive: opts.KeepAlive,
Michael Yang's avatar
Michael Yang committed
991
992
993
	}

	if err := client.Generate(ctx, &request, fn); err != nil {
994
		if errors.Is(err, context.Canceled) {
995
			return nil
996
		}
997
		return err
Patrick Devine's avatar
Patrick Devine committed
998
	}
999

1000
	if opts.Prompt != "" {
Michael Yang's avatar
Michael Yang committed
1001
1002
		fmt.Println()
		fmt.Println()
Patrick Devine's avatar
Patrick Devine committed
1003
	}
1004

1005
1006
1007
1008
	if !latest.Done {
		return nil
	}

Patrick Devine's avatar
Patrick Devine committed
1009
1010
	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
1011
		return err
Patrick Devine's avatar
Patrick Devine committed
1012
	}
Michael Yang's avatar
Michael Yang committed
1013

Patrick Devine's avatar
Patrick Devine committed
1014
1015
	if verbose {
		latest.Summary()
Michael Yang's avatar
Michael Yang committed
1016
	}
Michael Yang's avatar
Michael Yang committed
1017

Patrick Devine's avatar
Patrick Devine committed
1018
1019
1020
	ctx = context.WithValue(cmd.Context(), generateContextKey("context"), latest.Context)
	cmd.SetContext(ctx)

1021
	return nil
Michael Yang's avatar
Michael Yang committed
1022
1023
}

1024
func RunServer(_ *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1025
	if err := initializeKeypair(); err != nil {
1026
1027
1028
		return err
	}

Michael Yang's avatar
host  
Michael Yang committed
1029
	ln, err := net.Listen("tcp", envconfig.Host().Host)
1030
1031
1032
	if err != nil {
		return err
	}
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1033

1034
1035
1036
1037
1038
1039
	err = server.Serve(ln)
	if errors.Is(err, http.ErrServerClosed) {
		return nil
	}

	return err
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1040
1041
}

1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
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
1054
		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
1055
1056
1057
1058
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1059
		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
1060
1061
1062
1063
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1064
		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
1065
1066
1067
			return fmt.Errorf("could not create directory %w", err)
		}

Michael Yang's avatar
Michael Yang committed
1068
		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
1069
1070
1071
			return err
		}

Michael Yang's avatar
Michael Yang committed
1072
		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
1073
1074
1075
1076
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1077
		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
1078

Michael Yang's avatar
Michael Yang committed
1079
		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
1080
1081
1082
			return err
		}

Michael Yang's avatar
Michael Yang committed
1083
		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
1084
1085
1086
1087
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1088
func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1089
	client, err := api.ClientFromEnvironment()
1090
1091
1092
	if err != nil {
		return err
	}
Michael Yang's avatar
Michael Yang committed
1093
	if err := client.Heartbeat(cmd.Context()); err != nil {
1094
		if !strings.Contains(err.Error(), " refused") {
Bruce MacDonald's avatar
Bruce MacDonald committed
1095
1096
			return err
		}
1097
		if err := startApp(cmd.Context(), client); err != nil {
Michael Yang's avatar
lint  
Michael Yang committed
1098
			return errors.New("could not connect to ollama app, is it running?")
1099
1100
1101
1102
1103
		}
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1104
1105
1106
1107
1108
1109
1110
1111
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
1112
1113
1114
1115
1116
		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
1117
1118
	}

1119
	if serverVersion != version.Version {
Michael Yang's avatar
Michael Yang committed
1120
		fmt.Printf("Warning: client version is %s\n", version.Version)
1121
	}
Michael Yang's avatar
Michael Yang committed
1122
1123
}

1124
func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
1125
1126
1127
1128
1129
	if len(envs) == 0 {
		return
	}

	envUsage := `
1130
1131
Environment Variables:
`
1132
	for _, e := range envs {
1133
		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
1134
1135
1136
	}

	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
1137
1138
}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1139
1140
func NewCLI() *cobra.Command {
	log.SetFlags(log.LstdFlags | log.Lshortfile)
Michael Yang's avatar
Michael Yang committed
1141
	cobra.EnableCommandSorting = false
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1142

1143
	if runtime.GOOS == "windows" && term.IsTerminal(int(os.Stdout.Fd())) {
1144
		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
1145
1146
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1147
	rootCmd := &cobra.Command{
1148
1149
1150
1151
		Use:           "ollama",
		Short:         "Large language model runner",
		SilenceUsage:  true,
		SilenceErrors: true,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1152
1153
1154
		CompletionOptions: cobra.CompletionOptions{
			DisableDefaultCmd: true,
		},
Michael Yang's avatar
Michael Yang committed
1155
1156
1157
1158
1159
1160
1161
1162
		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
1163
1164
	}

Michael Yang's avatar
Michael Yang committed
1165
	rootCmd.Flags().BoolP("version", "v", false, "Show version information")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1166

1167
	createCmd := &cobra.Command{
1168
1169
		Use:     "create MODEL",
		Short:   "Create a model from a Modelfile",
Michael Yang's avatar
Michael Yang committed
1170
		Args:    cobra.ExactArgs(1),
1171
1172
		PreRunE: checkServerHeartbeat,
		RunE:    CreateHandler,
1173
1174
	}

1175
	createCmd.Flags().StringP("file", "f", "", "Name of the Modelfile (default \"Modelfile\"")
1176
	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_0)")
1177

Patrick Devine's avatar
Patrick Devine committed
1178
1179
1180
	showCmd := &cobra.Command{
		Use:     "show MODEL",
		Short:   "Show information for a model",
Michael Yang's avatar
Michael Yang committed
1181
		Args:    cobra.ExactArgs(1),
Patrick Devine's avatar
Patrick Devine committed
1182
1183
1184
1185
1186
1187
1188
1189
		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")
1190
	showCmd.Flags().Bool("system", false, "Show system message of a model")
Patrick Devine's avatar
Patrick Devine committed
1191

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1192
	runCmd := &cobra.Command{
1193
1194
1195
1196
1197
		Use:     "run MODEL [PROMPT]",
		Short:   "Run a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    RunHandler,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1198
1199
	}

1200
	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
1201
	runCmd.Flags().Bool("verbose", false, "Show timings for response")
1202
	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
1203
	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1204
	runCmd.Flags().String("format", "", "Response format (e.g. json)")
Patrick Devine's avatar
Patrick Devine committed
1205
1206
1207
1208
1209
1210
1211
1212
1213

	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
1214
1215
1216
1217
	serveCmd := &cobra.Command{
		Use:     "serve",
		Aliases: []string{"start"},
		Short:   "Start ollama",
Michael Yang's avatar
Michael Yang committed
1218
		Args:    cobra.ExactArgs(0),
Michael Yang's avatar
Michael Yang committed
1219
		RunE:    RunServer,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1220
1221
	}

1222
	pullCmd := &cobra.Command{
1223
1224
		Use:     "pull MODEL",
		Short:   "Pull a model from a registry",
Michael Yang's avatar
Michael Yang committed
1225
		Args:    cobra.ExactArgs(1),
1226
1227
		PreRunE: checkServerHeartbeat,
		RunE:    PullHandler,
1228
1229
	}

1230
1231
	pullCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1232
	pushCmd := &cobra.Command{
1233
1234
		Use:     "push MODEL",
		Short:   "Push a model to a registry",
Michael Yang's avatar
Michael Yang committed
1235
		Args:    cobra.ExactArgs(1),
1236
1237
		PreRunE: checkServerHeartbeat,
		RunE:    PushHandler,
1238
1239
	}

1240
1241
	pushCmd.Flags().Bool("insecure", false, "Use an insecure registry")

Patrick Devine's avatar
Patrick Devine committed
1242
	listCmd := &cobra.Command{
1243
		Use:     "list",
Patrick Devine's avatar
Patrick Devine committed
1244
		Aliases: []string{"ls"},
1245
		Short:   "List models",
1246
		PreRunE: checkServerHeartbeat,
1247
		RunE:    ListHandler,
1248
	}
1249
1250
1251
1252
1253
1254
1255
1256

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

Patrick Devine's avatar
Patrick Devine committed
1257
	copyCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1258
		Use:     "cp SOURCE DESTINATION",
1259
		Short:   "Copy a model",
Michael Yang's avatar
Michael Yang committed
1260
		Args:    cobra.ExactArgs(2),
1261
1262
		PreRunE: checkServerHeartbeat,
		RunE:    CopyHandler,
Patrick Devine's avatar
Patrick Devine committed
1263
1264
	}

1265
	deleteCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1266
		Use:     "rm MODEL [MODEL...]",
1267
1268
1269
1270
		Short:   "Remove a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    DeleteHandler,
Patrick Devine's avatar
Patrick Devine committed
1271
1272
	}

1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
	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:])
	})

1286
1287
1288
	envVars := envconfig.AsMap()

	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
1289

1290
1291
1292
1293
	for _, cmd := range []*cobra.Command{
		createCmd,
		showCmd,
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1294
		stopCmd,
1295
1296
1297
		pullCmd,
		pushCmd,
		listCmd,
1298
		psCmd,
1299
1300
		copyCmd,
		deleteCmd,
1301
		serveCmd,
1302
	} {
1303
1304
		switch cmd {
		case runCmd:
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
			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"],
1317
				envVars["OLLAMA_SCHED_SPREAD"],
1318
				envVars["OLLAMA_TMPDIR"],
1319
				envVars["OLLAMA_FLASH_ATTENTION"],
1320
				envVars["OLLAMA_KV_CACHE_TYPE"],
1321
				envVars["OLLAMA_LLM_LIBRARY"],
1322
				envVars["OLLAMA_GPU_OVERHEAD"],
1323
				envVars["OLLAMA_LOAD_TIMEOUT"],
1324
			})
1325
1326
1327
		default:
			appendEnvDocs(cmd, envs)
		}
1328
1329
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1330
1331
	rootCmd.AddCommand(
		serveCmd,
1332
		createCmd,
Patrick Devine's avatar
Patrick Devine committed
1333
		showCmd,
1334
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1335
		stopCmd,
1336
1337
		pullCmd,
		pushCmd,
Patrick Devine's avatar
Patrick Devine committed
1338
		listCmd,
1339
		psCmd,
Patrick Devine's avatar
Patrick Devine committed
1340
		copyCmd,
1341
		deleteCmd,
1342
		runnerCmd,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1343
1344
1345
1346
	)

	return rootCmd
}