cmd.go 30.8 KB
Newer Older
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1
2
3
package cmd

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

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

34
	"github.com/ollama/ollama/api"
35
	"github.com/ollama/ollama/envconfig"
36
	"github.com/ollama/ollama/format"
37
	"github.com/ollama/ollama/llama"
38
	"github.com/ollama/ollama/parser"
39
	"github.com/ollama/ollama/progress"
Jesse Gross's avatar
Jesse Gross committed
40
	"github.com/ollama/ollama/runner"
41
	"github.com/ollama/ollama/server"
42
	"github.com/ollama/ollama/types/model"
43
	"github.com/ollama/ollama/version"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
44
45
)

46
var errModelfileNotFound = errors.New("specified Modelfile wasn't found")
47
48

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

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

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

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

65
66
67
68
	return absName, nil
}

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

72
73
74
75
76
77
78
79
80
81
	var reader io.Reader

	filename, err := getModelfileName(cmd)
	if os.IsNotExist(err) {
		if filename == "" {
			reader = strings.NewReader("FROM .\n")
		} else {
			return errModelfileNotFound
		}
	} else if err != nil {
Michael Yang's avatar
Michael Yang committed
82
		return err
83
84
85
86
87
88
89
90
	} else {
		f, err := os.Open(filename)
		if err != nil {
			return err
		}

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

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

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

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

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

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

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

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

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

			bar.Set(resp.Completed)
		} else if status != resp.Status {
			spinner.Stop()

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

160
161
162
		return nil
	}

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

	return nil
}

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

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

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

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

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

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

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

221
222
223
224
225
226
227
228
229
type progressWriter struct {
	n atomic.Int64
}

func (w *progressWriter) Write(p []byte) (n int, err error) {
	w.n.Add(int64(len(p)))
	return len(p), nil
}

Patrick Devine's avatar
Patrick Devine committed
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
func loadOrUnloadModel(cmd *cobra.Command, opts *runOptions) error {
	p := progress.NewProgress(os.Stderr)
	defer p.StopAndClear()

	spinner := progress.NewSpinner("")
	p.Add("", spinner)

	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

	req := &api.GenerateRequest{
		Model:     opts.Model,
		KeepAlive: opts.KeepAlive,
	}

	return client.Generate(cmd.Context(), req, func(api.GenerateResponse) error { return nil })
}

func StopHandler(cmd *cobra.Command, args []string) error {
	opts := &runOptions{
		Model:     args[0],
		KeepAlive: &api.Duration{Duration: 0},
	}
	if err := loadOrUnloadModel(cmd, opts); err != nil {
		if strings.Contains(err.Error(), "not found") {
			return fmt.Errorf("couldn't find model \"%s\" to stop", args[0])
		}
259
		return err
Patrick Devine's avatar
Patrick Devine committed
260
261
262
263
	}
	return nil
}

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

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

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

279
280
281
282
283
284
285
286
287
288
289
290
	keepAlive, err := cmd.Flags().GetString("keepalive")
	if err != nil {
		return err
	}
	if keepAlive != "" {
		d, err := time.ParseDuration(keepAlive)
		if err != nil {
			return err
		}
		opts.KeepAlive = &api.Duration{Duration: d}
	}

291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
	prompts := args[1:]
	// prepend stdin to the prompt if provided
	if !term.IsTerminal(int(os.Stdin.Fd())) {
		in, err := io.ReadAll(os.Stdin)
		if err != nil {
			return err
		}

		prompts = append([]string{string(in)}, prompts...)
		opts.WordWrap = false
		interactive = false
	}
	opts.Prompt = strings.Join(prompts, " ")
	if len(prompts) > 0 {
		interactive = false
	}
307
308
309
310
	// Be quiet if we're redirecting to a pipe or file
	if !term.IsTerminal(int(os.Stdout.Fd())) {
		interactive = false
	}
311
312
313
314
315
316
317

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

318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
	// Fill out the rest of the options based on information about the
	// model.
	client, err := api.ClientFromEnvironment()
	if err != nil {
		return err
	}

	name := args[0]
	info, err := func() (*api.ShowResponse, error) {
		showReq := &api.ShowRequest{Name: name}
		info, err := client.Show(cmd.Context(), showReq)
		var se api.StatusError
		if errors.As(err, &se) && se.StatusCode == http.StatusNotFound {
			if err := PullHandler(cmd, []string{name}); err != nil {
				return nil, err
			}
			return client.Show(cmd.Context(), &api.ShowRequest{Name: name})
		}
		return info, err
	}()
	if err != nil {
		return err
340
341
	}

Jesse Gross's avatar
Jesse Gross committed
342
343
344
345
	// TODO(jessegross): We should either find another way to know if this is
	// a vision model or remove the logic. Also consider that other modalities will
	// need different behavior anyways.
	opts.MultiModal = len(info.ProjectorInfo) != 0 || envconfig.NewEngine()
346
347
348
	opts.ParentModel = info.Details.ParentModel

	if interactive {
Patrick Devine's avatar
Patrick Devine committed
349
		if err := loadOrUnloadModel(cmd, &opts); err != nil {
Michael Yang's avatar
Michael Yang committed
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
			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()
			}
		}

365
366
367
		return generateInteractive(cmd, opts)
	}
	return generate(cmd, opts)
Bruce MacDonald's avatar
Bruce MacDonald committed
368
369
}

370
func PushHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
371
	client, err := api.ClientFromEnvironment()
372
373
374
	if err != nil {
		return err
	}
375

376
377
378
379
380
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
381
382
383
384
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)
385
386
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
387

388
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
389
		if resp.Digest != "" {
390
391
392
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
393
394
395

			bar, ok := bars[resp.Digest]
			if !ok {
396
				bar = progress.NewBar(fmt.Sprintf("pushing %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
397
398
399
400
401
402
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
403
404
405
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
406
407
408
409
410
411

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

412
413
414
		return nil
	}

Michael Yang's avatar
Michael Yang committed
415
	request := api.PushRequest{Name: args[0], Insecure: insecure}
416
417

	n := model.ParseName(args[0])
Michael Yang's avatar
Michael Yang committed
418
	if err := client.Push(cmd.Context(), &request, fn); err != nil {
419
420
421
422
423
424
		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
425
426
427
		return err
	}

428
	p.Stop()
429
	spinner.Stop()
430
431
432
433
434
435
436
437

	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
438
	return nil
439
440
}

441
func ListHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
442
	client, err := api.ClientFromEnvironment()
443
444
445
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
446

Michael Yang's avatar
Michael Yang committed
447
	models, err := client.List(cmd.Context())
Patrick Devine's avatar
Patrick Devine committed
448
449
450
451
452
453
454
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
455
		if len(args) == 0 || strings.HasPrefix(strings.ToLower(m.Name), strings.ToLower(args[0])) {
456
			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
457
		}
Patrick Devine's avatar
Patrick Devine committed
458
459
460
	}

	table := tablewriter.NewWriter(os.Stdout)
Patrick Devine's avatar
Patrick Devine committed
461
	table.SetHeader([]string{"NAME", "ID", "SIZE", "MODIFIED"})
Patrick Devine's avatar
Patrick Devine committed
462
463
464
465
466
	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
467
	table.SetTablePadding("    ")
Patrick Devine's avatar
Patrick Devine committed
468
469
470
471
472
473
	table.AppendBulk(data)
	table.Render()

	return nil
}

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
499
500
501
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
502
503
504
505
506
507
508
509
510

			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})
511
512
513
514
515
516
517
518
519
520
		}
	}

	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
521
	table.SetTablePadding("    ")
522
523
524
525
526
527
	table.AppendBulk(data)
	table.Render()

	return nil
}

528
func DeleteHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
529
	client, err := api.ClientFromEnvironment()
530
531
532
	if err != nil {
		return err
	}
533

534
535
536
537
538
539
540
541
542
543
544
	// 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)
		}
	}

545
546
	for _, name := range args {
		req := api.DeleteRequest{Name: name}
Michael Yang's avatar
Michael Yang committed
547
		if err := client.Delete(cmd.Context(), &req); err != nil {
548
549
550
			return err
		}
		fmt.Printf("deleted '%s'\n", name)
551
552
553
554
	}
	return nil
}

Patrick Devine's avatar
Patrick Devine committed
555
func ShowHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
556
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
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
599
600
601
	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 {
602
		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
603
604
	}

605
606
607
608
609
	req := api.ShowRequest{Name: args[0]}
	resp, err := client.Show(cmd.Context(), &req)
	if err != nil {
		return err
	}
610

611
	if flagsSet == 1 {
612
613
614
615
616
617
618
619
		switch showType {
		case "license":
			fmt.Println(resp.License)
		case "modelfile":
			fmt.Println(resp.Modelfile)
		case "parameters":
			fmt.Println(resp.Parameters)
		case "system":
620
			fmt.Print(resp.System)
621
		case "template":
622
			fmt.Print(resp.Template)
623
624
625
		}

		return nil
Patrick Devine's avatar
Patrick Devine committed
626
627
	}

Michael Yang's avatar
Michael Yang committed
628
	return showInfo(resp, os.Stdout)
629
630
}

Michael Yang's avatar
Michael Yang committed
631
632
633
634
635
636
637
638
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("    ")
639

Michael Yang's avatar
Michael Yang committed
640
641
642
		switch header {
		case "Template", "System", "License":
			table.SetColWidth(100)
643
644
		}

Michael Yang's avatar
Michael Yang committed
645
646
647
		table.AppendBulk(rows())
		table.Render()
		fmt.Fprintln(w)
Patrick Devine's avatar
Patrick Devine committed
648
649
	}

Michael Yang's avatar
Michael Yang committed
650
651
652
653
654
655
656
657
658
659
660
661
662
663
	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
	})
664

Michael Yang's avatar
Michael Yang committed
665
666
667
668
669
670
671
672
673
	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
		})
674
675
	}

Michael Yang's avatar
Michael Yang committed
676
677
678
679
680
681
682
683
684
685
	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
		})
686
687
	}

Michael Yang's avatar
Michael Yang committed
688
689
690
691
692
	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)})
693
694
			}
		}
Michael Yang's avatar
Michael Yang committed
695
		return
696
697
	}

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

Michael Yang's avatar
Michael Yang committed
704
705
706
707
	if resp.License != "" {
		tableRender("License", func() [][]string {
			return head(resp.License, 2)
		})
708
	}
Michael Yang's avatar
Michael Yang committed
709
710

	return nil
711
712
}

Patrick Devine's avatar
Patrick Devine committed
713
func CopyHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
714
	client, err := api.ClientFromEnvironment()
715
716
717
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
718
719

	req := api.CopyRequest{Source: args[0], Destination: args[1]}
Michael Yang's avatar
Michael Yang committed
720
	if err := client.Copy(cmd.Context(), &req); err != nil {
Patrick Devine's avatar
Patrick Devine committed
721
722
723
724
725
726
		return err
	}
	fmt.Printf("copied '%s' to '%s'\n", args[0], args[1])
	return nil
}

727
func PullHandler(cmd *cobra.Command, args []string) error {
728
729
730
731
732
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
733
	client, err := api.ClientFromEnvironment()
734
735
736
	if err != nil {
		return err
	}
737

Michael Yang's avatar
Michael Yang committed
738
739
740
741
742
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

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

743
744
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
745

746
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
747
		if resp.Digest != "" {
748
749
750
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
751
752
753

			bar, ok := bars[resp.Digest]
			if !ok {
754
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
755
756
757
758
759
760
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
761
762
763
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
764
765
766
767
768
769

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

770
771
		return nil
	}
772

Michael Yang's avatar
Michael Yang committed
773
	request := api.PullRequest{Name: args[0], Insecure: insecure}
Michael Yang's avatar
Michael Yang committed
774
	if err := client.Pull(cmd.Context(), &request, fn); err != nil {
Michael Yang's avatar
Michael Yang committed
775
776
777
778
		return err
	}

	return nil
Michael Yang's avatar
Michael Yang committed
779
780
}

781
782
type generateContextKey string

783
type runOptions struct {
784
785
786
787
788
789
790
791
792
793
	Model       string
	ParentModel string
	Prompt      string
	Messages    []api.Message
	WordWrap    bool
	Format      string
	System      string
	Images      []api.ImageData
	Options     map[string]interface{}
	MultiModal  bool
794
	KeepAlive   *api.Duration
795
796
}

797
798
799
800
801
802
803
804
805
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
806
807
			if state.lineLength+1 > termWidth-5 {
				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
808
809
810
811
812
813
814
					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
815
816
				a := runewidth.StringWidth(state.wordBuffer)
				if a > 0 {
817
					fmt.Printf("\x1b[%dD", a)
818
819
				}
				fmt.Printf("\x1b[K\n")
820
				fmt.Printf("%s%c", state.wordBuffer, ch)
821
822
823
				chWidth := runewidth.RuneWidth(ch)

				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
824
825
			} else {
				fmt.Print(string(ch))
826
827
828
829
				state.lineLength += runewidth.RuneWidth(ch)
				if runewidth.RuneWidth(ch) >= 2 {
					state.wordBuffer = ""
					continue
Josh Yan's avatar
Josh Yan committed
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
889
890
891

				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
	}

892
893
894
895
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

896
897
898
	req := &api.ChatRequest{
		Model:    opts.Model,
		Messages: opts.Messages,
899
		Format:   json.RawMessage(opts.Format),
900
901
902
		Options:  opts.Options,
	}

903
904
905
906
	if opts.KeepAlive != nil {
		req.KeepAlive = opts.KeepAlive
	}

907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
	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
932
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
933
	if err != nil {
934
		return err
Patrick Devine's avatar
Patrick Devine committed
935
	}
Michael Yang's avatar
Michael Yang committed
936

Michael Yang's avatar
Michael Yang committed
937
	p := progress.NewProgress(os.Stderr)
938
	defer p.StopAndClear()
939

Michael Yang's avatar
Michael Yang committed
940
941
942
	spinner := progress.NewSpinner("")
	p.Add("", spinner)

943
944
945
946
947
948
949
	var latest api.GenerateResponse

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

Michael Yang's avatar
Michael Yang committed
950
	ctx, cancel := context.WithCancel(cmd.Context())
951
952
953
954
955
956
957
958
959
960
	defer cancel()

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

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

961
	var state *displayResponseState = &displayResponseState{}
962

963
	fn := func(response api.GenerateResponse) error {
Michael Yang's avatar
Michael Yang committed
964
		p.StopAndClear()
965

Patrick Devine's avatar
Patrick Devine committed
966
		latest = response
967
		content := response.Response
968

969
		displayResponse(content, opts.WordWrap, state)
970

Patrick Devine's avatar
Patrick Devine committed
971
972
		return nil
	}
973

974
975
976
977
978
979
980
	if opts.MultiModal {
		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
		if err != nil {
			return err
		}
	}

981
982
983
984
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

Michael Yang's avatar
Michael Yang committed
985
	request := api.GenerateRequest{
986
987
988
989
		Model:     opts.Model,
		Prompt:    opts.Prompt,
		Context:   generateContext,
		Images:    opts.Images,
990
		Format:    json.RawMessage(opts.Format),
991
992
993
		System:    opts.System,
		Options:   opts.Options,
		KeepAlive: opts.KeepAlive,
Michael Yang's avatar
Michael Yang committed
994
995
996
	}

	if err := client.Generate(ctx, &request, fn); err != nil {
997
		if errors.Is(err, context.Canceled) {
998
			return nil
999
		}
1000
		return err
Patrick Devine's avatar
Patrick Devine committed
1001
	}
1002

1003
	if opts.Prompt != "" {
Michael Yang's avatar
Michael Yang committed
1004
1005
		fmt.Println()
		fmt.Println()
Patrick Devine's avatar
Patrick Devine committed
1006
	}
1007

1008
1009
1010
1011
	if !latest.Done {
		return nil
	}

Patrick Devine's avatar
Patrick Devine committed
1012
1013
	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
1014
		return err
Patrick Devine's avatar
Patrick Devine committed
1015
	}
Michael Yang's avatar
Michael Yang committed
1016

Patrick Devine's avatar
Patrick Devine committed
1017
1018
	if verbose {
		latest.Summary()
Michael Yang's avatar
Michael Yang committed
1019
	}
Michael Yang's avatar
Michael Yang committed
1020

Patrick Devine's avatar
Patrick Devine committed
1021
1022
1023
	ctx = context.WithValue(cmd.Context(), generateContextKey("context"), latest.Context)
	cmd.SetContext(ctx)

1024
	return nil
Michael Yang's avatar
Michael Yang committed
1025
1026
}

1027
func RunServer(_ *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1028
	if err := initializeKeypair(); err != nil {
1029
1030
1031
		return err
	}

Michael Yang's avatar
host  
Michael Yang committed
1032
	ln, err := net.Listen("tcp", envconfig.Host().Host)
1033
1034
1035
	if err != nil {
		return err
	}
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1036

1037
1038
1039
1040
1041
1042
	err = server.Serve(ln)
	if errors.Is(err, http.ErrServerClosed) {
		return nil
	}

	return err
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1043
1044
}

1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
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
1057
		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
1058
1059
1060
1061
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1062
		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
1063
1064
1065
1066
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1067
		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
1068
1069
1070
			return fmt.Errorf("could not create directory %w", err)
		}

Michael Yang's avatar
Michael Yang committed
1071
		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
1072
1073
1074
			return err
		}

Michael Yang's avatar
Michael Yang committed
1075
		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
1076
1077
1078
1079
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1080
		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
1081

Michael Yang's avatar
Michael Yang committed
1082
		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
1083
1084
1085
			return err
		}

Michael Yang's avatar
Michael Yang committed
1086
		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
1087
1088
1089
1090
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1091
func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1092
	client, err := api.ClientFromEnvironment()
1093
1094
1095
	if err != nil {
		return err
	}
Michael Yang's avatar
Michael Yang committed
1096
	if err := client.Heartbeat(cmd.Context()); err != nil {
1097
		if !strings.Contains(err.Error(), " refused") {
Bruce MacDonald's avatar
Bruce MacDonald committed
1098
1099
			return err
		}
1100
		if err := startApp(cmd.Context(), client); err != nil {
Michael Yang's avatar
lint  
Michael Yang committed
1101
			return errors.New("could not connect to ollama app, is it running?")
1102
1103
1104
1105
1106
		}
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1107
1108
1109
1110
1111
1112
1113
1114
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
1115
1116
1117
1118
1119
		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
1120
1121
	}

1122
	if serverVersion != version.Version {
Michael Yang's avatar
Michael Yang committed
1123
		fmt.Printf("Warning: client version is %s\n", version.Version)
1124
	}
Michael Yang's avatar
Michael Yang committed
1125
1126
}

1127
func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
1128
1129
1130
1131
1132
	if len(envs) == 0 {
		return
	}

	envUsage := `
1133
1134
Environment Variables:
`
1135
	for _, e := range envs {
1136
		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
1137
1138
1139
	}

	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
1140
1141
}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1142
1143
func NewCLI() *cobra.Command {
	log.SetFlags(log.LstdFlags | log.Lshortfile)
Michael Yang's avatar
Michael Yang committed
1144
	cobra.EnableCommandSorting = false
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1145

1146
	if runtime.GOOS == "windows" && term.IsTerminal(int(os.Stdout.Fd())) {
1147
		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
1148
1149
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1150
	rootCmd := &cobra.Command{
1151
1152
1153
1154
		Use:           "ollama",
		Short:         "Large language model runner",
		SilenceUsage:  true,
		SilenceErrors: true,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1155
1156
1157
		CompletionOptions: cobra.CompletionOptions{
			DisableDefaultCmd: true,
		},
Michael Yang's avatar
Michael Yang committed
1158
1159
1160
1161
1162
1163
1164
1165
		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
1166
1167
	}

Michael Yang's avatar
Michael Yang committed
1168
	rootCmd.Flags().BoolP("version", "v", false, "Show version information")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1169

1170
	createCmd := &cobra.Command{
1171
1172
		Use:     "create MODEL",
		Short:   "Create a model from a Modelfile",
Michael Yang's avatar
Michael Yang committed
1173
		Args:    cobra.ExactArgs(1),
1174
1175
		PreRunE: checkServerHeartbeat,
		RunE:    CreateHandler,
1176
1177
	}

1178
	createCmd.Flags().StringP("file", "f", "", "Name of the Modelfile (default \"Modelfile\"")
1179
	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_0)")
1180

Patrick Devine's avatar
Patrick Devine committed
1181
1182
1183
	showCmd := &cobra.Command{
		Use:     "show MODEL",
		Short:   "Show information for a model",
Michael Yang's avatar
Michael Yang committed
1184
		Args:    cobra.ExactArgs(1),
Patrick Devine's avatar
Patrick Devine committed
1185
1186
1187
1188
1189
1190
1191
1192
		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")
1193
	showCmd.Flags().Bool("system", false, "Show system message of a model")
Patrick Devine's avatar
Patrick Devine committed
1194

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1195
	runCmd := &cobra.Command{
1196
1197
1198
1199
1200
		Use:     "run MODEL [PROMPT]",
		Short:   "Run a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    RunHandler,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1201
1202
	}

1203
	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
1204
	runCmd.Flags().Bool("verbose", false, "Show timings for response")
1205
	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
1206
	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1207
	runCmd.Flags().String("format", "", "Response format (e.g. json)")
Patrick Devine's avatar
Patrick Devine committed
1208
1209
1210
1211
1212
1213
1214
1215
1216

	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
1217
1218
1219
1220
	serveCmd := &cobra.Command{
		Use:     "serve",
		Aliases: []string{"start"},
		Short:   "Start ollama",
Michael Yang's avatar
Michael Yang committed
1221
		Args:    cobra.ExactArgs(0),
Michael Yang's avatar
Michael Yang committed
1222
		RunE:    RunServer,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1223
1224
	}

1225
	pullCmd := &cobra.Command{
1226
1227
		Use:     "pull MODEL",
		Short:   "Pull a model from a registry",
Michael Yang's avatar
Michael Yang committed
1228
		Args:    cobra.ExactArgs(1),
1229
1230
		PreRunE: checkServerHeartbeat,
		RunE:    PullHandler,
1231
1232
	}

1233
1234
	pullCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1235
	pushCmd := &cobra.Command{
1236
1237
		Use:     "push MODEL",
		Short:   "Push a model to a registry",
Michael Yang's avatar
Michael Yang committed
1238
		Args:    cobra.ExactArgs(1),
1239
1240
		PreRunE: checkServerHeartbeat,
		RunE:    PushHandler,
1241
1242
	}

1243
1244
	pushCmd.Flags().Bool("insecure", false, "Use an insecure registry")

Patrick Devine's avatar
Patrick Devine committed
1245
	listCmd := &cobra.Command{
1246
		Use:     "list",
Patrick Devine's avatar
Patrick Devine committed
1247
		Aliases: []string{"ls"},
1248
		Short:   "List models",
1249
		PreRunE: checkServerHeartbeat,
1250
		RunE:    ListHandler,
1251
	}
1252
1253
1254
1255
1256
1257
1258
1259

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

Patrick Devine's avatar
Patrick Devine committed
1260
	copyCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1261
		Use:     "cp SOURCE DESTINATION",
1262
		Short:   "Copy a model",
Michael Yang's avatar
Michael Yang committed
1263
		Args:    cobra.ExactArgs(2),
1264
1265
		PreRunE: checkServerHeartbeat,
		RunE:    CopyHandler,
Patrick Devine's avatar
Patrick Devine committed
1266
1267
	}

1268
	deleteCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1269
		Use:     "rm MODEL [MODEL...]",
1270
1271
1272
1273
		Short:   "Remove a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    DeleteHandler,
Patrick Devine's avatar
Patrick Devine committed
1274
1275
	}

1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
	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:])
	})

1289
1290
1291
	envVars := envconfig.AsMap()

	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
1292

1293
1294
1295
1296
	for _, cmd := range []*cobra.Command{
		createCmd,
		showCmd,
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1297
		stopCmd,
1298
1299
1300
		pullCmd,
		pushCmd,
		listCmd,
1301
		psCmd,
1302
1303
		copyCmd,
		deleteCmd,
1304
		serveCmd,
1305
	} {
1306
1307
		switch cmd {
		case runCmd:
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
			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"],
1320
				envVars["OLLAMA_SCHED_SPREAD"],
1321
				envVars["OLLAMA_TMPDIR"],
1322
				envVars["OLLAMA_FLASH_ATTENTION"],
1323
				envVars["OLLAMA_KV_CACHE_TYPE"],
1324
				envVars["OLLAMA_LLM_LIBRARY"],
1325
				envVars["OLLAMA_GPU_OVERHEAD"],
1326
				envVars["OLLAMA_LOAD_TIMEOUT"],
1327
			})
1328
1329
1330
		default:
			appendEnvDocs(cmd, envs)
		}
1331
1332
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1333
1334
	rootCmd.AddCommand(
		serveCmd,
1335
		createCmd,
Patrick Devine's avatar
Patrick Devine committed
1336
		showCmd,
1337
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1338
		stopCmd,
1339
1340
		pullCmd,
		pushCmd,
Patrick Devine's avatar
Patrick Devine committed
1341
		listCmd,
1342
		psCmd,
Patrick Devine's avatar
Patrick Devine committed
1343
		copyCmd,
1344
		deleteCmd,
1345
		runnerCmd,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1346
1347
1348
1349
	)

	return rootCmd
}