cmd.go 34.7 KB
Newer Older
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1
2
3
package cmd

import (
4
	"archive/zip"
Michael Yang's avatar
Michael Yang committed
5
	"bufio"
Michael Yang's avatar
Michael Yang committed
6
	"bytes"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
7
	"context"
8
9
	"crypto/ed25519"
	"crypto/rand"
Michael Yang's avatar
Michael Yang committed
10
	"crypto/sha256"
11
	"encoding/json"
12
	"encoding/pem"
Michael Yang's avatar
Michael Yang committed
13
	"errors"
Bruce MacDonald's avatar
Bruce MacDonald committed
14
	"fmt"
Michael Yang's avatar
Michael Yang committed
15
	"io"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
16
	"log"
17
	"math"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
18
	"net"
19
	"net/http"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
20
	"os"
21
	"os/signal"
22
	"path/filepath"
23
	"runtime"
Michael Yang's avatar
Michael Yang committed
24
	"strconv"
Michael Yang's avatar
Michael Yang committed
25
	"strings"
26
	"sync/atomic"
27
	"syscall"
Michael Yang's avatar
Michael Yang committed
28
	"time"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
29

30
	"github.com/containerd/console"
31
	"github.com/mattn/go-runewidth"
Patrick Devine's avatar
Patrick Devine committed
32
	"github.com/olekukonko/tablewriter"
Michael Yang's avatar
Michael Yang committed
33
	"github.com/spf13/cobra"
34
	"golang.org/x/crypto/ssh"
35
	"golang.org/x/term"
Michael Yang's avatar
Michael Yang committed
36

37
	"github.com/ollama/ollama/api"
38
	"github.com/ollama/ollama/envconfig"
39
	"github.com/ollama/ollama/format"
40
41
	"github.com/ollama/ollama/llama"
	"github.com/ollama/ollama/llama/runner"
42
	"github.com/ollama/ollama/parser"
43
44
	"github.com/ollama/ollama/progress"
	"github.com/ollama/ollama/server"
45
	"github.com/ollama/ollama/types/model"
46
	"github.com/ollama/ollama/version"
Jeffrey Morgan's avatar
Jeffrey Morgan committed
47
48
)

49
50
51
52
53
54
55
56
57
58
59
60
61
62
var (
	errModelNotFound     = errors.New("no Modelfile or safetensors files found")
	errModelfileNotFound = errors.New("specified Modelfile wasn't found")
)

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

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

	absName, err := filepath.Abs(filename)
63
	if err != nil {
64
		return "", err
65
66
	}

67
	_, err = os.Stat(absName)
68
	if err != nil {
69
		return fn, err
70
	}
71

72
73
74
75
	return absName, nil
}

func CreateHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
76
77
78
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

79
80
81
82
83
84
85
86
87
88
	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
89
		return err
90
91
92
93
94
95
96
97
	} else {
		f, err := os.Open(filename)
		if err != nil {
			return err
		}

		reader = f
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
98
99
	}

100
	modelfile, err := parser.ParseFile(reader)
Michael Yang's avatar
Michael Yang committed
101
102
103
104
105
106
107
108
109
	if err != nil {
		return err
	}

	home, err := os.UserHomeDir()
	if err != nil {
		return err
	}

110
111
	status := "transferring model data"
	spinner := progress.NewSpinner(status)
112
	p.Add(status, spinner)
113
	defer p.Stop()
114

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

Michael Yang's avatar
Michael Yang committed
120
121
	for i := range modelfile.Commands {
		switch modelfile.Commands[i].Name {
Michael Yang's avatar
Michael Yang committed
122
		case "model", "adapter":
Michael Yang's avatar
Michael Yang committed
123
			path := modelfile.Commands[i].Args
Michael Yang's avatar
Michael Yang committed
124
125
126
127
128
129
			if path == "~" {
				path = home
			} else if strings.HasPrefix(path, "~/") {
				path = filepath.Join(home, path[2:])
			}

130
131
132
133
			if !filepath.IsAbs(path) {
				path = filepath.Join(filepath.Dir(filename), path)
			}

134
			fi, err := os.Stat(path)
Michael Yang's avatar
Michael Yang committed
135
			if errors.Is(err, os.ErrNotExist) && modelfile.Commands[i].Name == "model" {
Michael Yang's avatar
Michael Yang committed
136
				continue
Michael Yang's avatar
Michael Yang committed
137
138
139
140
			} else if err != nil {
				return err
			}

141
			if fi.IsDir() {
Michael Yang's avatar
Michael Yang committed
142
143
144
				// this is likely a safetensors or pytorch directory
				// TODO make this work w/ adapters
				tempfile, err := tempZipFiles(path)
145
146
147
				if err != nil {
					return err
				}
Michael Yang's avatar
Michael Yang committed
148
				defer os.RemoveAll(tempfile)
149

Michael Yang's avatar
Michael Yang committed
150
				path = tempfile
Michael Yang's avatar
Michael Yang committed
151
152
			}

153
			digest, err := createBlob(cmd, client, path, spinner)
154
			if err != nil {
Michael Yang's avatar
Michael Yang committed
155
156
157
				return err
			}

Michael Yang's avatar
Michael Yang committed
158
			modelfile.Commands[i].Args = "@" + digest
Michael Yang's avatar
Michael Yang committed
159
160
		}
	}
Michael Yang's avatar
Michael Yang committed
161

Michael Yang's avatar
Michael Yang committed
162
	bars := make(map[string]*progress.Bar)
163
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
164
165
166
167
168
		if resp.Digest != "" {
			spinner.Stop()

			bar, ok := bars[resp.Digest]
			if !ok {
169
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
170
171
172
173
174
175
176
177
178
179
180
181
182
				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)
		}

183
184
185
		return nil
	}

186
	quantize, _ := cmd.Flags().GetString("quantize")
Michael Yang's avatar
Michael Yang committed
187

188
	request := api.CreateRequest{Name: args[0], Modelfile: modelfile.String(), Quantize: quantize}
Michael Yang's avatar
Michael Yang committed
189
	if err := client.Create(cmd.Context(), &request, fn); err != nil {
190
191
192
193
194
195
		return err
	}

	return nil
}

Michael Yang's avatar
Michael Yang committed
196
197
198
199
200
201
202
func tempZipFiles(path string) (string, error) {
	tempfile, err := os.CreateTemp("", "ollama-tf")
	if err != nil {
		return "", err
	}
	defer tempfile.Close()

Michael Yang's avatar
Michael Yang committed
203
204
	detectContentType := func(path string) (string, error) {
		f, err := os.Open(path)
Michael Yang's avatar
Michael Yang committed
205
206
207
		if err != nil {
			return "", err
		}
Michael Yang's avatar
Michael Yang committed
208
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
209

Michael Yang's avatar
Michael Yang committed
210
211
		var b bytes.Buffer
		b.Grow(512)
Michael Yang's avatar
Michael Yang committed
212

Michael Yang's avatar
Michael Yang committed
213
214
215
216
217
218
		if _, err := io.CopyN(&b, f, 512); err != nil && !errors.Is(err, io.EOF) {
			return "", err
		}

		contentType, _, _ := strings.Cut(http.DetectContentType(b.Bytes()), ";")
		return contentType, nil
Michael Yang's avatar
Michael Yang committed
219
220
	}

Michael Yang's avatar
Michael Yang committed
221
222
223
224
225
226
227
228
229
230
231
	glob := func(pattern, contentType string) ([]string, error) {
		matches, err := filepath.Glob(pattern)
		if err != nil {
			return nil, err
		}

		for _, safetensor := range matches {
			if ct, err := detectContentType(safetensor); err != nil {
				return nil, err
			} else if ct != contentType {
				return nil, fmt.Errorf("invalid content type: expected %s for %s", ct, safetensor)
Michael Yang's avatar
Michael Yang committed
232
			}
Michael Yang's avatar
Michael Yang committed
233
234
235
236
237
238
239
240
241
242
		}

		return matches, nil
	}

	var files []string
	if st, _ := glob(filepath.Join(path, "model*.safetensors"), "application/octet-stream"); len(st) > 0 {
		// safetensors files might be unresolved git lfs references; skip if they are
		// covers model-x-of-y.safetensors, model.fp32-x-of-y.safetensors, model.safetensors
		files = append(files, st...)
243
244
245
246
247
248
	} else if st, _ := glob(filepath.Join(path, "adapters.safetensors"), "application/octet-stream"); len(st) > 0 {
		// covers adapters.safetensors
		files = append(files, st...)
	} else if st, _ := glob(filepath.Join(path, "adapter_model.safetensors"), "application/octet-stream"); len(st) > 0 {
		// covers adapter_model.safetensors
		files = append(files, st...)
Michael Yang's avatar
Michael Yang committed
249
250
251
252
	} else if pt, _ := glob(filepath.Join(path, "pytorch_model*.bin"), "application/zip"); len(pt) > 0 {
		// pytorch files might also be unresolved git lfs references; skip if they are
		// covers pytorch_model-x-of-y.bin, pytorch_model.fp32-x-of-y.bin, pytorch_model.bin
		files = append(files, pt...)
Patrick Devine's avatar
Patrick Devine committed
253
	} else if pt, _ := glob(filepath.Join(path, "consolidated*.pth"), "application/zip"); len(pt) > 0 {
Michael Yang's avatar
Michael Yang committed
254
255
256
257
		// pytorch files might also be unresolved git lfs references; skip if they are
		// covers consolidated.x.pth, consolidated.pth
		files = append(files, pt...)
	} else {
258
		return "", errModelNotFound
Michael Yang's avatar
Michael Yang committed
259
260
261
262
263
264
265
266
267
	}

	// add configuration files, json files are detected as text/plain
	js, err := glob(filepath.Join(path, "*.json"), "text/plain")
	if err != nil {
		return "", err
	}
	files = append(files, js...)

Michael Yang's avatar
Michael Yang committed
268
269
270
271
272
273
274
275
	// bert models require a nested config.json
	// TODO(mxyng): merge this with the glob above
	js, err = glob(filepath.Join(path, "**/*.json"), "text/plain")
	if err != nil {
		return "", err
	}
	files = append(files, js...)

Michael Yang's avatar
Michael Yang committed
276
277
278
279
280
281
282
283
284
	if tks, _ := glob(filepath.Join(path, "tokenizer.model"), "application/octet-stream"); len(tks) > 0 {
		// add tokenizer.model if it exists, tokenizer.json is automatically picked up by the previous glob
		// tokenizer.model might be a unresolved git lfs reference; error if it is
		files = append(files, tks...)
	} else if tks, _ := glob(filepath.Join(path, "**/tokenizer.model"), "text/plain"); len(tks) > 0 {
		// some times tokenizer.model is in a subdirectory (e.g. meta-llama/Meta-Llama-3-8B)
		files = append(files, tks...)
	}

285
286
287
	zipfile := zip.NewWriter(tempfile)
	defer zipfile.Close()

Michael Yang's avatar
Michael Yang committed
288
289
290
	for _, file := range files {
		f, err := os.Open(file)
		if err != nil {
Michael Yang's avatar
Michael Yang committed
291
292
			return "", err
		}
Michael Yang's avatar
Michael Yang committed
293
		defer f.Close()
Michael Yang's avatar
Michael Yang committed
294
295
296
297
298
299

		fi, err := f.Stat()
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
300
		zfi, err := zip.FileInfoHeader(fi)
Michael Yang's avatar
Michael Yang committed
301
302
303
304
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
305
306
307
308
309
		zfi.Name, err = filepath.Rel(path, file)
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
310
		zf, err := zipfile.CreateHeader(zfi)
Michael Yang's avatar
Michael Yang committed
311
312
313
314
		if err != nil {
			return "", err
		}

Michael Yang's avatar
Michael Yang committed
315
		if _, err := io.Copy(zf, f); err != nil {
Michael Yang's avatar
Michael Yang committed
316
317
318
319
320
321
322
			return "", err
		}
	}

	return tempfile.Name(), nil
}

323
func createBlob(cmd *cobra.Command, client *api.Client, path string, spinner *progress.Spinner) (string, error) {
324
325
326
327
328
329
	bin, err := os.Open(path)
	if err != nil {
		return "", err
	}
	defer bin.Close()

330
331
332
333
334
335
336
	// Get file info to retrieve the size
	fileInfo, err := bin.Stat()
	if err != nil {
		return "", err
	}
	fileSize := fileInfo.Size()

337
338
339
340
	hash := sha256.New()
	if _, err := io.Copy(hash, bin); err != nil {
		return "", err
	}
341
342
343
344

	if _, err := bin.Seek(0, io.SeekStart); err != nil {
		return "", err
	}
345

346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
	var pw progressWriter
	status := "transferring model data 0%"
	spinner.SetMessage(status)

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

	go func() {
		ticker := time.NewTicker(60 * time.Millisecond)
		defer ticker.Stop()
		for {
			select {
			case <-ticker.C:
				spinner.SetMessage(fmt.Sprintf("transferring model data %d%%", int(100*pw.n.Load()/fileSize)))
			case <-done:
				spinner.SetMessage("transferring model data 100%")
				return
			}
		}
	}()

367
	digest := fmt.Sprintf("sha256:%x", hash.Sum(nil))
368
	if err = client.CreateBlob(cmd.Context(), digest, io.TeeReader(bin, &pw)); err != nil {
369
370
371
372
373
		return "", err
	}
	return digest, nil
}

374
375
376
377
378
379
380
381
382
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
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
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
}

416
func RunHandler(cmd *cobra.Command, args []string) error {
417
418
419
	interactive := true

	opts := runOptions{
420
421
422
		Model:    args[0],
		WordWrap: os.Getenv("TERM") == "xterm-256color",
		Options:  map[string]interface{}{},
423
424
425
426
427
428
429
430
	}

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

431
432
433
434
435
436
437
438
439
440
441
442
	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}
	}

443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
	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
	}
459
460
461
462
	// Be quiet if we're redirecting to a pipe or file
	if !term.IsTerminal(int(os.Stdout.Fd())) {
		interactive = false
	}
463
464
465
466
467
468
469

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

470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
	// 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
492
493
	}

494
	opts.MultiModal = len(info.ProjectorInfo) != 0
495
496
497
	opts.ParentModel = info.Details.ParentModel

	if interactive {
Patrick Devine's avatar
Patrick Devine committed
498
		if err := loadOrUnloadModel(cmd, &opts); err != nil {
Michael Yang's avatar
Michael Yang committed
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
			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()
			}
		}

514
515
516
		return generateInteractive(cmd, opts)
	}
	return generate(cmd, opts)
Bruce MacDonald's avatar
Bruce MacDonald committed
517
518
}

519
func PushHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
520
	client, err := api.ClientFromEnvironment()
521
522
523
	if err != nil {
		return err
	}
524

525
526
527
528
529
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
530
531
532
533
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

	bars := make(map[string]*progress.Bar)
534
535
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
536

537
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
538
		if resp.Digest != "" {
539
540
541
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
542
543
544

			bar, ok := bars[resp.Digest]
			if !ok {
545
				bar = progress.NewBar(fmt.Sprintf("pushing %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
546
547
548
549
550
551
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
552
553
554
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
555
556
557
558
559
560

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

561
562
563
		return nil
	}

Michael Yang's avatar
Michael Yang committed
564
	request := api.PushRequest{Name: args[0], Insecure: insecure}
565
566

	n := model.ParseName(args[0])
Michael Yang's avatar
Michael Yang committed
567
	if err := client.Push(cmd.Context(), &request, fn); err != nil {
568
569
570
571
572
573
		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
574
575
576
		return err
	}

577
	p.Stop()
578
	spinner.Stop()
579
580
581
582
583
584
585
586

	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
587
	return nil
588
589
}

590
func ListHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
591
	client, err := api.ClientFromEnvironment()
592
593
594
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
595

Michael Yang's avatar
Michael Yang committed
596
	models, err := client.List(cmd.Context())
Patrick Devine's avatar
Patrick Devine committed
597
598
599
600
601
602
603
	if err != nil {
		return err
	}

	var data [][]string

	for _, m := range models.Models {
Michael Yang's avatar
Michael Yang committed
604
		if len(args) == 0 || strings.HasPrefix(m.Name, args[0]) {
605
			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
606
		}
Patrick Devine's avatar
Patrick Devine committed
607
608
609
	}

	table := tablewriter.NewWriter(os.Stdout)
Patrick Devine's avatar
Patrick Devine committed
610
	table.SetHeader([]string{"NAME", "ID", "SIZE", "MODIFIED"})
Patrick Devine's avatar
Patrick Devine committed
611
612
613
614
615
	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
616
	table.SetTablePadding("    ")
Patrick Devine's avatar
Patrick Devine committed
617
618
619
620
621
622
	table.AppendBulk(data)
	table.Render()

	return nil
}

623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
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
651
652
653
654
655
656
657
658
659

			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})
660
661
662
663
664
665
666
667
668
669
		}
	}

	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
670
	table.SetTablePadding("    ")
671
672
673
674
675
676
	table.AppendBulk(data)
	table.Render()

	return nil
}

677
func DeleteHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
678
	client, err := api.ClientFromEnvironment()
679
680
681
	if err != nil {
		return err
	}
682

683
684
685
686
687
688
689
690
691
692
693
	// 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)
		}
	}

694
695
	for _, name := range args {
		req := api.DeleteRequest{Name: name}
Michael Yang's avatar
Michael Yang committed
696
		if err := client.Delete(cmd.Context(), &req); err != nil {
697
698
699
			return err
		}
		fmt.Printf("deleted '%s'\n", name)
700
701
702
703
	}
	return nil
}

Patrick Devine's avatar
Patrick Devine committed
704
func ShowHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
705
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
	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 {
751
		return errors.New("only one of '--license', '--modelfile', '--parameters', '--system', or '--template' can be specified")
752
753
	}

754
755
756
757
758
	req := api.ShowRequest{Name: args[0]}
	resp, err := client.Show(cmd.Context(), &req)
	if err != nil {
		return err
	}
759

760
	if flagsSet == 1 {
761
762
763
764
765
766
767
768
		switch showType {
		case "license":
			fmt.Println(resp.License)
		case "modelfile":
			fmt.Println(resp.Modelfile)
		case "parameters":
			fmt.Println(resp.Parameters)
		case "system":
769
			fmt.Print(resp.System)
770
		case "template":
771
			fmt.Print(resp.Template)
772
773
774
		}

		return nil
Patrick Devine's avatar
Patrick Devine committed
775
776
	}

Michael Yang's avatar
Michael Yang committed
777
	return showInfo(resp, os.Stdout)
778
779
}

Michael Yang's avatar
Michael Yang committed
780
781
782
783
784
785
786
787
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("    ")
788

Michael Yang's avatar
Michael Yang committed
789
790
791
		switch header {
		case "Template", "System", "License":
			table.SetColWidth(100)
792
793
		}

Michael Yang's avatar
Michael Yang committed
794
795
796
		table.AppendBulk(rows())
		table.Render()
		fmt.Fprintln(w)
Patrick Devine's avatar
Patrick Devine committed
797
798
	}

Michael Yang's avatar
Michael Yang committed
799
800
801
802
803
804
805
806
807
808
809
810
811
812
	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
	})
813

Michael Yang's avatar
Michael Yang committed
814
815
816
817
818
819
820
821
822
	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
		})
823
824
	}

Michael Yang's avatar
Michael Yang committed
825
826
827
828
829
830
831
832
833
834
	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
		})
835
836
	}

Michael Yang's avatar
Michael Yang committed
837
838
839
840
841
	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)})
842
843
			}
		}
Michael Yang's avatar
Michael Yang committed
844
		return
845
846
	}

Michael Yang's avatar
Michael Yang committed
847
848
849
850
851
	if resp.System != "" {
		tableRender("System", func() [][]string {
			return head(resp.System, 2)
		})
	}
852

Michael Yang's avatar
Michael Yang committed
853
854
855
856
	if resp.License != "" {
		tableRender("License", func() [][]string {
			return head(resp.License, 2)
		})
857
	}
Michael Yang's avatar
Michael Yang committed
858
859

	return nil
860
861
}

Patrick Devine's avatar
Patrick Devine committed
862
func CopyHandler(cmd *cobra.Command, args []string) error {
Michael Yang's avatar
Michael Yang committed
863
	client, err := api.ClientFromEnvironment()
864
865
866
	if err != nil {
		return err
	}
Patrick Devine's avatar
Patrick Devine committed
867
868

	req := api.CopyRequest{Source: args[0], Destination: args[1]}
Michael Yang's avatar
Michael Yang committed
869
	if err := client.Copy(cmd.Context(), &req); err != nil {
Patrick Devine's avatar
Patrick Devine committed
870
871
872
873
874
875
		return err
	}
	fmt.Printf("copied '%s' to '%s'\n", args[0], args[1])
	return nil
}

876
func PullHandler(cmd *cobra.Command, args []string) error {
877
878
879
880
881
	insecure, err := cmd.Flags().GetBool("insecure")
	if err != nil {
		return err
	}

Michael Yang's avatar
Michael Yang committed
882
	client, err := api.ClientFromEnvironment()
883
884
885
	if err != nil {
		return err
	}
886

Michael Yang's avatar
Michael Yang committed
887
888
889
890
891
	p := progress.NewProgress(os.Stderr)
	defer p.Stop()

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

892
893
	var status string
	var spinner *progress.Spinner
Michael Yang's avatar
Michael Yang committed
894

895
	fn := func(resp api.ProgressResponse) error {
Michael Yang's avatar
Michael Yang committed
896
		if resp.Digest != "" {
897
898
899
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
900
901
902

			bar, ok := bars[resp.Digest]
			if !ok {
903
				bar = progress.NewBar(fmt.Sprintf("pulling %s...", resp.Digest[7:19]), resp.Total, resp.Completed)
Michael Yang's avatar
Michael Yang committed
904
905
906
907
908
909
				bars[resp.Digest] = bar
				p.Add(resp.Digest, bar)
			}

			bar.Set(resp.Completed)
		} else if status != resp.Status {
910
911
912
			if spinner != nil {
				spinner.Stop()
			}
Michael Yang's avatar
Michael Yang committed
913
914
915
916
917
918

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

919
920
		return nil
	}
921

Michael Yang's avatar
Michael Yang committed
922
	request := api.PullRequest{Name: args[0], Insecure: insecure}
Michael Yang's avatar
Michael Yang committed
923
	if err := client.Pull(cmd.Context(), &request, fn); err != nil {
Michael Yang's avatar
Michael Yang committed
924
925
926
927
		return err
	}

	return nil
Michael Yang's avatar
Michael Yang committed
928
929
}

930
931
type generateContextKey string

932
type runOptions struct {
933
934
935
936
937
938
939
940
941
942
	Model       string
	ParentModel string
	Prompt      string
	Messages    []api.Message
	WordWrap    bool
	Format      string
	System      string
	Images      []api.ImageData
	Options     map[string]interface{}
	MultiModal  bool
943
	KeepAlive   *api.Duration
944
945
}

946
947
948
949
950
951
952
953
954
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
955
956
			if state.lineLength+1 > termWidth-5 {
				if runewidth.StringWidth(state.wordBuffer) > termWidth-10 {
957
958
959
960
961
962
963
					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
964
965
				a := runewidth.StringWidth(state.wordBuffer)
				if a > 0 {
966
					fmt.Printf("\x1b[%dD", a)
967
968
				}
				fmt.Printf("\x1b[K\n")
969
				fmt.Printf("%s%c", state.wordBuffer, ch)
970
971
972
				chWidth := runewidth.RuneWidth(ch)

				state.lineLength = runewidth.StringWidth(state.wordBuffer) + chWidth
973
974
			} else {
				fmt.Print(string(ch))
975
976
977
978
				state.lineLength += runewidth.RuneWidth(ch)
				if runewidth.RuneWidth(ch) >= 2 {
					state.wordBuffer = ""
					continue
Josh Yan's avatar
Josh Yan committed
979
				}
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040

				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
	}

1041
1042
1043
1044
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

1045
1046
1047
	req := &api.ChatRequest{
		Model:    opts.Model,
		Messages: opts.Messages,
1048
		Format:   json.RawMessage(opts.Format),
1049
1050
1051
		Options:  opts.Options,
	}

1052
1053
1054
1055
	if opts.KeepAlive != nil {
		req.KeepAlive = opts.KeepAlive
	}

1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
	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
1081
	client, err := api.ClientFromEnvironment()
Patrick Devine's avatar
Patrick Devine committed
1082
	if err != nil {
1083
		return err
Patrick Devine's avatar
Patrick Devine committed
1084
	}
Michael Yang's avatar
Michael Yang committed
1085

Michael Yang's avatar
Michael Yang committed
1086
	p := progress.NewProgress(os.Stderr)
1087
	defer p.StopAndClear()
1088

Michael Yang's avatar
Michael Yang committed
1089
1090
1091
	spinner := progress.NewSpinner("")
	p.Add("", spinner)

1092
1093
1094
1095
1096
1097
1098
	var latest api.GenerateResponse

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

Michael Yang's avatar
Michael Yang committed
1099
	ctx, cancel := context.WithCancel(cmd.Context())
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
	defer cancel()

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

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

1110
	var state *displayResponseState = &displayResponseState{}
1111

1112
	fn := func(response api.GenerateResponse) error {
Michael Yang's avatar
Michael Yang committed
1113
		p.StopAndClear()
1114

Patrick Devine's avatar
Patrick Devine committed
1115
		latest = response
1116
		content := response.Response
1117

1118
		displayResponse(content, opts.WordWrap, state)
1119

Patrick Devine's avatar
Patrick Devine committed
1120
1121
		return nil
	}
1122

1123
1124
1125
1126
1127
1128
1129
	if opts.MultiModal {
		opts.Prompt, opts.Images, err = extractFileData(opts.Prompt)
		if err != nil {
			return err
		}
	}

1130
1131
1132
1133
	if opts.Format == "json" {
		opts.Format = `"` + opts.Format + `"`
	}

Michael Yang's avatar
Michael Yang committed
1134
	request := api.GenerateRequest{
1135
1136
1137
1138
		Model:     opts.Model,
		Prompt:    opts.Prompt,
		Context:   generateContext,
		Images:    opts.Images,
1139
		Format:    json.RawMessage(opts.Format),
1140
1141
1142
		System:    opts.System,
		Options:   opts.Options,
		KeepAlive: opts.KeepAlive,
Michael Yang's avatar
Michael Yang committed
1143
1144
1145
	}

	if err := client.Generate(ctx, &request, fn); err != nil {
1146
		if errors.Is(err, context.Canceled) {
1147
			return nil
1148
		}
1149
		return err
Patrick Devine's avatar
Patrick Devine committed
1150
	}
1151

1152
	if opts.Prompt != "" {
Michael Yang's avatar
Michael Yang committed
1153
1154
		fmt.Println()
		fmt.Println()
Patrick Devine's avatar
Patrick Devine committed
1155
	}
1156

1157
1158
1159
1160
	if !latest.Done {
		return nil
	}

Patrick Devine's avatar
Patrick Devine committed
1161
1162
	verbose, err := cmd.Flags().GetBool("verbose")
	if err != nil {
1163
		return err
Patrick Devine's avatar
Patrick Devine committed
1164
	}
Michael Yang's avatar
Michael Yang committed
1165

Patrick Devine's avatar
Patrick Devine committed
1166
1167
	if verbose {
		latest.Summary()
Michael Yang's avatar
Michael Yang committed
1168
	}
Michael Yang's avatar
Michael Yang committed
1169

Patrick Devine's avatar
Patrick Devine committed
1170
1171
1172
	ctx = context.WithValue(cmd.Context(), generateContextKey("context"), latest.Context)
	cmd.SetContext(ctx)

1173
	return nil
Michael Yang's avatar
Michael Yang committed
1174
1175
}

1176
func RunServer(_ *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1177
	if err := initializeKeypair(); err != nil {
1178
1179
1180
		return err
	}

Michael Yang's avatar
host  
Michael Yang committed
1181
	ln, err := net.Listen("tcp", envconfig.Host().Host)
1182
1183
1184
	if err != nil {
		return err
	}
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1185

1186
1187
1188
1189
1190
1191
	err = server.Serve(ln)
	if errors.Is(err, http.ErrServerClosed) {
		return nil
	}

	return err
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1192
1193
}

1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
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
1206
		cryptoPublicKey, cryptoPrivateKey, err := ed25519.GenerateKey(rand.Reader)
1207
1208
1209
1210
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1211
		privateKeyBytes, err := ssh.MarshalPrivateKey(cryptoPrivateKey, "")
1212
1213
1214
1215
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1216
		if err := os.MkdirAll(filepath.Dir(privKeyPath), 0o755); err != nil {
1217
1218
1219
			return fmt.Errorf("could not create directory %w", err)
		}

Michael Yang's avatar
Michael Yang committed
1220
		if err := os.WriteFile(privKeyPath, pem.EncodeToMemory(privateKeyBytes), 0o600); err != nil {
1221
1222
1223
			return err
		}

Michael Yang's avatar
Michael Yang committed
1224
		sshPublicKey, err := ssh.NewPublicKey(cryptoPublicKey)
1225
1226
1227
1228
		if err != nil {
			return err
		}

Michael Yang's avatar
Michael Yang committed
1229
		publicKeyBytes := ssh.MarshalAuthorizedKey(sshPublicKey)
1230

Michael Yang's avatar
Michael Yang committed
1231
		if err := os.WriteFile(pubKeyPath, publicKeyBytes, 0o644); err != nil {
1232
1233
1234
			return err
		}

Michael Yang's avatar
Michael Yang committed
1235
		fmt.Printf("Your new public key is: \n\n%s\n", publicKeyBytes)
1236
1237
1238
1239
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1240
func checkServerHeartbeat(cmd *cobra.Command, _ []string) error {
Michael Yang's avatar
Michael Yang committed
1241
	client, err := api.ClientFromEnvironment()
1242
1243
1244
	if err != nil {
		return err
	}
Michael Yang's avatar
Michael Yang committed
1245
	if err := client.Heartbeat(cmd.Context()); err != nil {
1246
		if !strings.Contains(err.Error(), " refused") {
Bruce MacDonald's avatar
Bruce MacDonald committed
1247
1248
			return err
		}
1249
		if err := startApp(cmd.Context(), client); err != nil {
Michael Yang's avatar
lint  
Michael Yang committed
1250
			return errors.New("could not connect to ollama app, is it running?")
1251
1252
1253
1254
1255
		}
	}
	return nil
}

Michael Yang's avatar
Michael Yang committed
1256
1257
1258
1259
1260
1261
1262
1263
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
1264
1265
1266
1267
1268
		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
1269
1270
	}

1271
	if serverVersion != version.Version {
Michael Yang's avatar
Michael Yang committed
1272
		fmt.Printf("Warning: client version is %s\n", version.Version)
1273
	}
Michael Yang's avatar
Michael Yang committed
1274
1275
}

1276
func appendEnvDocs(cmd *cobra.Command, envs []envconfig.EnvVar) {
1277
1278
1279
1280
1281
	if len(envs) == 0 {
		return
	}

	envUsage := `
1282
1283
Environment Variables:
`
1284
	for _, e := range envs {
1285
		envUsage += fmt.Sprintf("      %-24s   %s\n", e.Name, e.Description)
1286
1287
1288
	}

	cmd.SetUsageTemplate(cmd.UsageTemplate() + envUsage)
1289
1290
}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1291
1292
func NewCLI() *cobra.Command {
	log.SetFlags(log.LstdFlags | log.Lshortfile)
Michael Yang's avatar
Michael Yang committed
1293
	cobra.EnableCommandSorting = false
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1294

1295
	if runtime.GOOS == "windows" && term.IsTerminal(int(os.Stdout.Fd())) {
1296
		console.ConsoleFromFile(os.Stdin) //nolint:errcheck
1297
1298
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1299
	rootCmd := &cobra.Command{
1300
1301
1302
1303
		Use:           "ollama",
		Short:         "Large language model runner",
		SilenceUsage:  true,
		SilenceErrors: true,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1304
1305
1306
		CompletionOptions: cobra.CompletionOptions{
			DisableDefaultCmd: true,
		},
Michael Yang's avatar
Michael Yang committed
1307
1308
1309
1310
1311
1312
1313
1314
		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
1315
1316
	}

Michael Yang's avatar
Michael Yang committed
1317
	rootCmd.Flags().BoolP("version", "v", false, "Show version information")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1318

1319
	createCmd := &cobra.Command{
1320
1321
		Use:     "create MODEL",
		Short:   "Create a model from a Modelfile",
Michael Yang's avatar
Michael Yang committed
1322
		Args:    cobra.ExactArgs(1),
1323
1324
		PreRunE: checkServerHeartbeat,
		RunE:    CreateHandler,
1325
1326
	}

1327
	createCmd.Flags().StringP("file", "f", "", "Name of the Modelfile (default \"Modelfile\"")
1328
	createCmd.Flags().StringP("quantize", "q", "", "Quantize model to this level (e.g. q4_0)")
1329

Patrick Devine's avatar
Patrick Devine committed
1330
1331
1332
	showCmd := &cobra.Command{
		Use:     "show MODEL",
		Short:   "Show information for a model",
Michael Yang's avatar
Michael Yang committed
1333
		Args:    cobra.ExactArgs(1),
Patrick Devine's avatar
Patrick Devine committed
1334
1335
1336
1337
1338
1339
1340
1341
		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")
1342
	showCmd.Flags().Bool("system", false, "Show system message of a model")
Patrick Devine's avatar
Patrick Devine committed
1343

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1344
	runCmd := &cobra.Command{
1345
1346
1347
1348
1349
		Use:     "run MODEL [PROMPT]",
		Short:   "Run a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    RunHandler,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1350
1351
	}

1352
	runCmd.Flags().String("keepalive", "", "Duration to keep a model loaded (e.g. 5m)")
1353
	runCmd.Flags().Bool("verbose", false, "Show timings for response")
1354
	runCmd.Flags().Bool("insecure", false, "Use an insecure registry")
1355
	runCmd.Flags().Bool("nowordwrap", false, "Don't wrap words to the next line automatically")
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1356
	runCmd.Flags().String("format", "", "Response format (e.g. json)")
Patrick Devine's avatar
Patrick Devine committed
1357
1358
1359
1360
1361
1362
1363
1364
1365

	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
1366
1367
1368
1369
	serveCmd := &cobra.Command{
		Use:     "serve",
		Aliases: []string{"start"},
		Short:   "Start ollama",
Michael Yang's avatar
Michael Yang committed
1370
		Args:    cobra.ExactArgs(0),
Michael Yang's avatar
Michael Yang committed
1371
		RunE:    RunServer,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1372
1373
	}

1374
	pullCmd := &cobra.Command{
1375
1376
		Use:     "pull MODEL",
		Short:   "Pull a model from a registry",
Michael Yang's avatar
Michael Yang committed
1377
		Args:    cobra.ExactArgs(1),
1378
1379
		PreRunE: checkServerHeartbeat,
		RunE:    PullHandler,
1380
1381
	}

1382
1383
	pullCmd.Flags().Bool("insecure", false, "Use an insecure registry")

1384
	pushCmd := &cobra.Command{
1385
1386
		Use:     "push MODEL",
		Short:   "Push a model to a registry",
Michael Yang's avatar
Michael Yang committed
1387
		Args:    cobra.ExactArgs(1),
1388
1389
		PreRunE: checkServerHeartbeat,
		RunE:    PushHandler,
1390
1391
	}

1392
1393
	pushCmd.Flags().Bool("insecure", false, "Use an insecure registry")

Patrick Devine's avatar
Patrick Devine committed
1394
	listCmd := &cobra.Command{
1395
		Use:     "list",
Patrick Devine's avatar
Patrick Devine committed
1396
		Aliases: []string{"ls"},
1397
		Short:   "List models",
1398
		PreRunE: checkServerHeartbeat,
1399
		RunE:    ListHandler,
1400
	}
1401
1402
1403
1404
1405
1406
1407
1408

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

Patrick Devine's avatar
Patrick Devine committed
1409
	copyCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1410
		Use:     "cp SOURCE DESTINATION",
1411
		Short:   "Copy a model",
Michael Yang's avatar
Michael Yang committed
1412
		Args:    cobra.ExactArgs(2),
1413
1414
		PreRunE: checkServerHeartbeat,
		RunE:    CopyHandler,
Patrick Devine's avatar
Patrick Devine committed
1415
1416
	}

1417
	deleteCmd := &cobra.Command{
Michael Yang's avatar
Michael Yang committed
1418
		Use:     "rm MODEL [MODEL...]",
1419
1420
1421
1422
		Short:   "Remove a model",
		Args:    cobra.MinimumNArgs(1),
		PreRunE: checkServerHeartbeat,
		RunE:    DeleteHandler,
Patrick Devine's avatar
Patrick Devine committed
1423
1424
	}

1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
	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:])
	})

1438
1439
1440
	envVars := envconfig.AsMap()

	envs := []envconfig.EnvVar{envVars["OLLAMA_HOST"]}
1441

1442
1443
1444
1445
	for _, cmd := range []*cobra.Command{
		createCmd,
		showCmd,
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1446
		stopCmd,
1447
1448
1449
		pullCmd,
		pushCmd,
		listCmd,
1450
		psCmd,
1451
1452
		copyCmd,
		deleteCmd,
1453
		serveCmd,
1454
	} {
1455
1456
		switch cmd {
		case runCmd:
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
			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"],
1469
				envVars["OLLAMA_SCHED_SPREAD"],
1470
				envVars["OLLAMA_TMPDIR"],
1471
				envVars["OLLAMA_FLASH_ATTENTION"],
1472
				envVars["OLLAMA_KV_CACHE_TYPE"],
1473
				envVars["OLLAMA_LLM_LIBRARY"],
1474
				envVars["OLLAMA_GPU_OVERHEAD"],
1475
				envVars["OLLAMA_LOAD_TIMEOUT"],
1476
			})
1477
1478
1479
		default:
			appendEnvDocs(cmd, envs)
		}
1480
1481
	}

Jeffrey Morgan's avatar
Jeffrey Morgan committed
1482
1483
	rootCmd.AddCommand(
		serveCmd,
1484
		createCmd,
Patrick Devine's avatar
Patrick Devine committed
1485
		showCmd,
1486
		runCmd,
Patrick Devine's avatar
Patrick Devine committed
1487
		stopCmd,
1488
1489
		pullCmd,
		pushCmd,
Patrick Devine's avatar
Patrick Devine committed
1490
		listCmd,
1491
		psCmd,
Patrick Devine's avatar
Patrick Devine committed
1492
		copyCmd,
1493
		deleteCmd,
1494
		runnerCmd,
Jeffrey Morgan's avatar
Jeffrey Morgan committed
1495
1496
1497
1498
	)

	return rootCmd
}