main.py 34.4 KB
Newer Older
Timothy J. Baek's avatar
Timothy J. Baek committed
1
2
3
4
5
6
7
8
9
10
11
from fastapi import (
    FastAPI,
    Request,
    Response,
    HTTPException,
    Depends,
    status,
    UploadFile,
    File,
    BackgroundTasks,
)
12
13
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import StreamingResponse
Timothy J. Baek's avatar
Timothy J. Baek committed
14
from fastapi.concurrency import run_in_threadpool
Timothy J. Baek's avatar
Timothy J. Baek committed
15

Timothy J. Baek's avatar
Timothy J. Baek committed
16
from pydantic import BaseModel, ConfigDict
17

Timothy J. Baek's avatar
Timothy J. Baek committed
18
import os
Timothy J. Baek's avatar
Timothy J. Baek committed
19
import copy
20
import random
Timothy J. Baek's avatar
Timothy J. Baek committed
21
22
import requests
import json
23
import uuid
24
25
import aiohttp
import asyncio
Timothy J. Baek's avatar
Timothy J. Baek committed
26
27
28
from urllib.parse import urlparse
from typing import Optional, List, Union

Timothy J. Baek's avatar
Timothy J. Baek committed
29

30
31
from apps.web.models.users import Users
from constants import ERROR_MESSAGES
32
from utils.utils import decode_token, get_current_user, get_admin_user
Timothy J. Baek's avatar
Timothy J. Baek committed
33
from utils.misc import calculate_sha256
Timothy J. Baek's avatar
Timothy J. Baek committed
34

Timothy J. Baek's avatar
Timothy J. Baek committed
35
36

from config import OLLAMA_BASE_URLS, MODEL_FILTER_ENABLED, MODEL_FILTER_LIST, UPLOAD_DIR
37
38


39
40
41
42
43
44
45
46
app = FastAPI()
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)
Timothy J. Baek's avatar
Timothy J. Baek committed
47

Timothy J. Baek's avatar
Timothy J. Baek committed
48

49
50
app.state.MODEL_FILTER_ENABLED = MODEL_FILTER_ENABLED
app.state.MODEL_FILTER_LIST = MODEL_FILTER_LIST
Timothy J. Baek's avatar
Timothy J. Baek committed
51

52
app.state.OLLAMA_BASE_URLS = OLLAMA_BASE_URLS
53
app.state.MODELS = {}
Timothy J. Baek's avatar
Timothy J. Baek committed
54
55


56
57
58
REQUEST_POOL = []


Timothy J. Baek's avatar
Timothy J. Baek committed
59
60
61
62
63
# TODO: Implement a more intelligent load balancing mechanism for distributing requests among multiple backend instances.
# Current implementation uses a simple round-robin approach (random.choice). Consider incorporating algorithms like weighted round-robin,
# least connections, or least response time for better resource utilization and performance optimization.


64
65
66
67
68
69
70
71
72
73
74
75
76
77
@app.middleware("http")
async def check_url(request: Request, call_next):
    if len(app.state.MODELS) == 0:
        await get_all_models()
    else:
        pass

    response = await call_next(request)
    return response


@app.get("/urls")
async def get_ollama_api_urls(user=Depends(get_admin_user)):
    return {"OLLAMA_BASE_URLS": app.state.OLLAMA_BASE_URLS}
78

Timothy J. Baek's avatar
Timothy J. Baek committed
79

80
class UrlUpdateForm(BaseModel):
81
    urls: List[str]
82
83


84
@app.post("/urls/update")
Timothy J. Baek's avatar
Timothy J. Baek committed
85
async def update_ollama_api_url(form_data: UrlUpdateForm, user=Depends(get_admin_user)):
86
87
88
89
    app.state.OLLAMA_BASE_URLS = form_data.urls

    print(app.state.OLLAMA_BASE_URLS)
    return {"OLLAMA_BASE_URLS": app.state.OLLAMA_BASE_URLS}
Timothy J. Baek's avatar
Timothy J. Baek committed
90
91


92
93
94
95
96
97
98
99
100
101
@app.get("/cancel/{request_id}")
async def cancel_ollama_request(request_id: str, user=Depends(get_current_user)):
    if user:
        if request_id in REQUEST_POOL:
            REQUEST_POOL.remove(request_id)
        return True
    else:
        raise HTTPException(status_code=401, detail=ERROR_MESSAGES.ACCESS_PROHIBITED)


102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
async def fetch_url(url):
    try:
        async with aiohttp.ClientSession() as session:
            async with session.get(url) as response:
                return await response.json()
    except Exception as e:
        # Handle connection error here
        print(f"Connection error: {e}")
        return None


def merge_models_lists(model_lists):
    merged_models = {}

    for idx, model_list in enumerate(model_lists):
Timothy J. Baek's avatar
Timothy J. Baek committed
117
118
119
120
121
122
123
124
        if model_list is not None:
            for model in model_list:
                digest = model["digest"]
                if digest not in merged_models:
                    model["urls"] = [idx]
                    merged_models[digest] = model
                else:
                    merged_models[digest]["urls"].append(idx)
125
126
127
128
129
130
131
132
133
134
135
136
137
138

    return list(merged_models.values())


# user=Depends(get_current_user)


async def get_all_models():
    print("get_all_models")
    tasks = [fetch_url(f"{url}/api/tags") for url in app.state.OLLAMA_BASE_URLS]
    responses = await asyncio.gather(*tasks)

    models = {
        "models": merge_models_lists(
Timothy J. Baek's avatar
Timothy J. Baek committed
139
            map(lambda response: response["models"] if response else None, responses)
140
141
        )
    }
Timothy J. Baek's avatar
Timothy J. Baek committed
142

143
144
145
146
147
148
149
150
151
152
153
    app.state.MODELS = {model["model"]: model for model in models["models"]}

    return models


@app.get("/api/tags")
@app.get("/api/tags/{url_idx}")
async def get_ollama_tags(
    url_idx: Optional[int] = None, user=Depends(get_current_user)
):
    if url_idx == None:
Timothy J. Baek's avatar
Timothy J. Baek committed
154
        models = await get_all_models()
155

Timothy J. Baek's avatar
Timothy J. Baek committed
156
157
        if app.state.MODEL_FILTER_ENABLED:
            if user.role == "user":
158
159
                models["models"] = list(
                    filter(
160
                        lambda model: model["name"] in app.state.MODEL_FILTER_LIST,
161
162
                        models["models"],
                    )
Timothy J. Baek's avatar
Timothy J. Baek committed
163
164
165
                )
                return models
        return models
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
    else:
        url = app.state.OLLAMA_BASE_URLS[url_idx]
        try:
            r = requests.request(method="GET", url=f"{url}/api/tags")
            r.raise_for_status()

            return r.json()
        except Exception as e:
            print(e)
            error_detail = "Open WebUI: Server Connection Error"
            if r is not None:
                try:
                    res = r.json()
                    if "error" in res:
                        error_detail = f"Ollama: {res['error']}"
                except:
                    error_detail = f"Ollama: {e}"

            raise HTTPException(
                status_code=r.status_code if r else 500,
                detail=error_detail,
            )


@app.get("/api/version")
@app.get("/api/version/{url_idx}")
async def get_ollama_versions(url_idx: Optional[int] = None):

    if url_idx == None:

        # returns lowest version
        tasks = [fetch_url(f"{url}/api/version") for url in app.state.OLLAMA_BASE_URLS]
        responses = await asyncio.gather(*tasks)
        responses = list(filter(lambda x: x is not None, responses))

Timothy J. Baek's avatar
Timothy J. Baek committed
201
202
203
204
        if len(responses) > 0:
            lowest_version = min(
                responses, key=lambda x: tuple(map(int, x["version"].split(".")))
            )
205

Timothy J. Baek's avatar
Timothy J. Baek committed
206
207
208
209
210
211
            return {"version": lowest_version["version"]}
        else:
            raise HTTPException(
                status_code=500,
                detail=ERROR_MESSAGES.OLLAMA_NOT_FOUND,
            )
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
    else:
        url = app.state.OLLAMA_BASE_URLS[url_idx]
        try:
            r = requests.request(method="GET", url=f"{url}/api/version")
            r.raise_for_status()

            return r.json()
        except Exception as e:
            print(e)
            error_detail = "Open WebUI: Server Connection Error"
            if r is not None:
                try:
                    res = r.json()
                    if "error" in res:
                        error_detail = f"Ollama: {res['error']}"
                except:
                    error_detail = f"Ollama: {e}"

            raise HTTPException(
                status_code=r.status_code if r else 500,
                detail=error_detail,
            )


class ModelNameForm(BaseModel):
    name: str


@app.post("/api/pull")
@app.post("/api/pull/{url_idx}")
async def pull_model(
    form_data: ModelNameForm, url_idx: int = 0, user=Depends(get_admin_user)
):
    url = app.state.OLLAMA_BASE_URLS[url_idx]
246
247
    print(url)

248
249
    r = None

250
251
    def get_request():
        nonlocal url
252
253
254
255
256
257
258
259
260
261
        nonlocal r
        try:

            def stream_content():
                for chunk in r.iter_content(chunk_size=8192):
                    yield chunk

            r = requests.request(
                method="POST",
                url=f"{url}/api/pull",
Timothy J. Baek's avatar
Timothy J. Baek committed
262
                data=form_data.model_dump_json(exclude_none=True).encode(),
263
264
265
266
267
268
269
270
271
272
273
274
275
276
                stream=True,
            )

            r.raise_for_status()

            return StreamingResponse(
                stream_content(),
                status_code=r.status_code,
                headers=dict(r.headers),
            )
        except Exception as e:
            raise e

    try:
277
        return await run_in_threadpool(get_request)
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
    except Exception as e:
        print(e)
        error_detail = "Open WebUI: Server Connection Error"
        if r is not None:
            try:
                res = r.json()
                if "error" in res:
                    error_detail = f"Ollama: {res['error']}"
            except:
                error_detail = f"Ollama: {e}"

        raise HTTPException(
            status_code=r.status_code if r else 500,
            detail=error_detail,
        )


class PushModelForm(BaseModel):
    name: str
    insecure: Optional[bool] = None
    stream: Optional[bool] = None


@app.delete("/api/push")
@app.delete("/api/push/{url_idx}")
async def push_model(
    form_data: PushModelForm,
    url_idx: Optional[int] = None,
    user=Depends(get_admin_user),
):
    if url_idx == None:
        if form_data.name in app.state.MODELS:
            url_idx = app.state.MODELS[form_data.name]["urls"][0]
        else:
            raise HTTPException(
                status_code=400,
Timothy J. Baek's avatar
Timothy J. Baek committed
314
                detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.name),
315
316
317
            )

    url = app.state.OLLAMA_BASE_URLS[url_idx]
318
    print(url)
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333

    r = None

    def get_request():
        nonlocal url
        nonlocal r
        try:

            def stream_content():
                for chunk in r.iter_content(chunk_size=8192):
                    yield chunk

            r = requests.request(
                method="POST",
                url=f"{url}/api/push",
Timothy J. Baek's avatar
Timothy J. Baek committed
334
                data=form_data.model_dump_json(exclude_none=True).encode(),
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
            )

            r.raise_for_status()

            return StreamingResponse(
                stream_content(),
                status_code=r.status_code,
                headers=dict(r.headers),
            )
        except Exception as e:
            raise e

    try:
        return await run_in_threadpool(get_request)
    except Exception as e:
        print(e)
        error_detail = "Open WebUI: Server Connection Error"
        if r is not None:
            try:
                res = r.json()
                if "error" in res:
                    error_detail = f"Ollama: {res['error']}"
            except:
                error_detail = f"Ollama: {e}"

        raise HTTPException(
            status_code=r.status_code if r else 500,
            detail=error_detail,
        )


class CreateModelForm(BaseModel):
    name: str
    modelfile: Optional[str] = None
    stream: Optional[bool] = None
    path: Optional[str] = None


@app.post("/api/create")
@app.post("/api/create/{url_idx}")
async def create_model(
    form_data: CreateModelForm, url_idx: int = 0, user=Depends(get_admin_user)
):
    print(form_data)
    url = app.state.OLLAMA_BASE_URLS[url_idx]
380
381
    print(url)

382
383
384
385
386
387
388
389
390
391
392
393
394
395
    r = None

    def get_request():
        nonlocal url
        nonlocal r
        try:

            def stream_content():
                for chunk in r.iter_content(chunk_size=8192):
                    yield chunk

            r = requests.request(
                method="POST",
                url=f"{url}/api/create",
Timothy J. Baek's avatar
Timothy J. Baek committed
396
                data=form_data.model_dump_json(exclude_none=True).encode(),
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
                stream=True,
            )

            r.raise_for_status()

            print(r)

            return StreamingResponse(
                stream_content(),
                status_code=r.status_code,
                headers=dict(r.headers),
            )
        except Exception as e:
            raise e

    try:
        return await run_in_threadpool(get_request)
    except Exception as e:
        print(e)
        error_detail = "Open WebUI: Server Connection Error"
        if r is not None:
            try:
                res = r.json()
                if "error" in res:
                    error_detail = f"Ollama: {res['error']}"
            except:
                error_detail = f"Ollama: {e}"

        raise HTTPException(
            status_code=r.status_code if r else 500,
            detail=error_detail,
        )


class CopyModelForm(BaseModel):
    source: str
    destination: str


@app.post("/api/copy")
@app.post("/api/copy/{url_idx}")
async def copy_model(
    form_data: CopyModelForm,
    url_idx: Optional[int] = None,
    user=Depends(get_admin_user),
):
    if url_idx == None:
        if form_data.source in app.state.MODELS:
            url_idx = app.state.MODELS[form_data.source]["urls"][0]
        else:
            raise HTTPException(
                status_code=400,
Timothy J. Baek's avatar
Timothy J. Baek committed
449
                detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.source),
450
451
452
            )

    url = app.state.OLLAMA_BASE_URLS[url_idx]
453
    print(url)
454
455
456
457
458

    try:
        r = requests.request(
            method="POST",
            url=f"{url}/api/copy",
Timothy J. Baek's avatar
Timothy J. Baek committed
459
            data=form_data.model_dump_json(exclude_none=True).encode(),
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
        )
        r.raise_for_status()

        print(r.text)

        return True
    except Exception as e:
        print(e)
        error_detail = "Open WebUI: Server Connection Error"
        if r is not None:
            try:
                res = r.json()
                if "error" in res:
                    error_detail = f"Ollama: {res['error']}"
            except:
                error_detail = f"Ollama: {e}"

        raise HTTPException(
            status_code=r.status_code if r else 500,
            detail=error_detail,
        )


@app.delete("/api/delete")
@app.delete("/api/delete/{url_idx}")
async def delete_model(
    form_data: ModelNameForm,
    url_idx: Optional[int] = None,
    user=Depends(get_admin_user),
):
    if url_idx == None:
        if form_data.name in app.state.MODELS:
            url_idx = app.state.MODELS[form_data.name]["urls"][0]
        else:
            raise HTTPException(
                status_code=400,
Timothy J. Baek's avatar
Timothy J. Baek committed
496
                detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.name),
497
498
499
            )

    url = app.state.OLLAMA_BASE_URLS[url_idx]
500
    print(url)
501
502
503
504
505

    try:
        r = requests.request(
            method="DELETE",
            url=f"{url}/api/delete",
Timothy J. Baek's avatar
Timothy J. Baek committed
506
            data=form_data.model_dump_json(exclude_none=True).encode(),
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
        )
        r.raise_for_status()

        print(r.text)

        return True
    except Exception as e:
        print(e)
        error_detail = "Open WebUI: Server Connection Error"
        if r is not None:
            try:
                res = r.json()
                if "error" in res:
                    error_detail = f"Ollama: {res['error']}"
            except:
                error_detail = f"Ollama: {e}"

        raise HTTPException(
            status_code=r.status_code if r else 500,
            detail=error_detail,
        )


@app.post("/api/show")
async def show_model_info(form_data: ModelNameForm, user=Depends(get_current_user)):
    if form_data.name not in app.state.MODELS:
        raise HTTPException(
            status_code=400,
Timothy J. Baek's avatar
Timothy J. Baek committed
535
            detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.name),
536
537
538
539
        )

    url_idx = random.choice(app.state.MODELS[form_data.name]["urls"])
    url = app.state.OLLAMA_BASE_URLS[url_idx]
540
    print(url)
541
542
543
544
545

    try:
        r = requests.request(
            method="POST",
            url=f"{url}/api/show",
Timothy J. Baek's avatar
Timothy J. Baek committed
546
            data=form_data.model_dump_json(exclude_none=True).encode(),
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
        )
        r.raise_for_status()

        return r.json()
    except Exception as e:
        print(e)
        error_detail = "Open WebUI: Server Connection Error"
        if r is not None:
            try:
                res = r.json()
                if "error" in res:
                    error_detail = f"Ollama: {res['error']}"
            except:
                error_detail = f"Ollama: {e}"

        raise HTTPException(
            status_code=r.status_code if r else 500,
            detail=error_detail,
        )


class GenerateEmbeddingsForm(BaseModel):
    model: str
    prompt: str
    options: Optional[dict] = None
    keep_alive: Optional[Union[int, str]] = None


@app.post("/api/embeddings")
@app.post("/api/embeddings/{url_idx}")
async def generate_embeddings(
    form_data: GenerateEmbeddingsForm,
    url_idx: Optional[int] = None,
    user=Depends(get_current_user),
):
    if url_idx == None:
        if form_data.model in app.state.MODELS:
            url_idx = random.choice(app.state.MODELS[form_data.model]["urls"])
        else:
            raise HTTPException(
                status_code=400,
Timothy J. Baek's avatar
Timothy J. Baek committed
588
                detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.model),
589
590
591
            )

    url = app.state.OLLAMA_BASE_URLS[url_idx]
592
    print(url)
593
594
595
596
597

    try:
        r = requests.request(
            method="POST",
            url=f"{url}/api/embeddings",
Timothy J. Baek's avatar
Timothy J. Baek committed
598
            data=form_data.model_dump_json(exclude_none=True).encode(),
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
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
651
        )
        r.raise_for_status()

        return r.json()
    except Exception as e:
        print(e)
        error_detail = "Open WebUI: Server Connection Error"
        if r is not None:
            try:
                res = r.json()
                if "error" in res:
                    error_detail = f"Ollama: {res['error']}"
            except:
                error_detail = f"Ollama: {e}"

        raise HTTPException(
            status_code=r.status_code if r else 500,
            detail=error_detail,
        )


class GenerateCompletionForm(BaseModel):
    model: str
    prompt: str
    images: Optional[List[str]] = None
    format: Optional[str] = None
    options: Optional[dict] = None
    system: Optional[str] = None
    template: Optional[str] = None
    context: Optional[str] = None
    stream: Optional[bool] = True
    raw: Optional[bool] = None
    keep_alive: Optional[Union[int, str]] = None


@app.post("/api/generate")
@app.post("/api/generate/{url_idx}")
async def generate_completion(
    form_data: GenerateCompletionForm,
    url_idx: Optional[int] = None,
    user=Depends(get_current_user),
):

    if url_idx == None:
        if form_data.model in app.state.MODELS:
            url_idx = random.choice(app.state.MODELS[form_data.model]["urls"])
        else:
            raise HTTPException(
                status_code=400,
                detail="error_detail",
            )

    url = app.state.OLLAMA_BASE_URLS[url_idx]
652
    print(url)
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683

    r = None

    def get_request():
        nonlocal form_data
        nonlocal r

        request_id = str(uuid.uuid4())
        try:
            REQUEST_POOL.append(request_id)

            def stream_content():
                try:
                    if form_data.stream:
                        yield json.dumps({"id": request_id, "done": False}) + "\n"

                    for chunk in r.iter_content(chunk_size=8192):
                        if request_id in REQUEST_POOL:
                            yield chunk
                        else:
                            print("User: canceled request")
                            break
                finally:
                    if hasattr(r, "close"):
                        r.close()
                        if request_id in REQUEST_POOL:
                            REQUEST_POOL.remove(request_id)

            r = requests.request(
                method="POST",
                url=f"{url}/api/generate",
Timothy J. Baek's avatar
Timothy J. Baek committed
684
                data=form_data.model_dump_json(exclude_none=True).encode(),
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
                stream=True,
            )

            r.raise_for_status()

            return StreamingResponse(
                stream_content(),
                status_code=r.status_code,
                headers=dict(r.headers),
            )
        except Exception as e:
            raise e

    try:
        return await run_in_threadpool(get_request)
    except Exception as e:
        error_detail = "Open WebUI: Server Connection Error"
        if r is not None:
            try:
                res = r.json()
                if "error" in res:
                    error_detail = f"Ollama: {res['error']}"
            except:
                error_detail = f"Ollama: {e}"

        raise HTTPException(
            status_code=r.status_code if r else 500,
            detail=error_detail,
        )


class ChatMessage(BaseModel):
    role: str
    content: str
    images: Optional[List[str]] = None


class GenerateChatCompletionForm(BaseModel):
    model: str
    messages: List[ChatMessage]
    format: Optional[str] = None
    options: Optional[dict] = None
    template: Optional[str] = None
Timothy J. Baek's avatar
Timothy J. Baek committed
728
    stream: Optional[bool] = None
729
730
731
732
733
    keep_alive: Optional[Union[int, str]] = None


@app.post("/api/chat")
@app.post("/api/chat/{url_idx}")
Timothy J. Baek's avatar
Timothy J. Baek committed
734
async def generate_chat_completion(
735
736
737
738
739
740
741
742
743
744
745
    form_data: GenerateChatCompletionForm,
    url_idx: Optional[int] = None,
    user=Depends(get_current_user),
):

    if url_idx == None:
        if form_data.model in app.state.MODELS:
            url_idx = random.choice(app.state.MODELS[form_data.model]["urls"])
        else:
            raise HTTPException(
                status_code=400,
Timothy J. Baek's avatar
Timothy J. Baek committed
746
                detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.model),
747
748
749
            )

    url = app.state.OLLAMA_BASE_URLS[url_idx]
750
    print(url)
751
752
753

    r = None

Timothy J. Baek's avatar
Timothy J. Baek committed
754
    print(form_data.model_dump_json(exclude_none=True).encode())
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783

    def get_request():
        nonlocal form_data
        nonlocal r

        request_id = str(uuid.uuid4())
        try:
            REQUEST_POOL.append(request_id)

            def stream_content():
                try:
                    if form_data.stream:
                        yield json.dumps({"id": request_id, "done": False}) + "\n"

                    for chunk in r.iter_content(chunk_size=8192):
                        if request_id in REQUEST_POOL:
                            yield chunk
                        else:
                            print("User: canceled request")
                            break
                finally:
                    if hasattr(r, "close"):
                        r.close()
                        if request_id in REQUEST_POOL:
                            REQUEST_POOL.remove(request_id)

            r = requests.request(
                method="POST",
                url=f"{url}/api/chat",
Timothy J. Baek's avatar
Timothy J. Baek committed
784
                data=form_data.model_dump_json(exclude_none=True).encode(),
785
786
787
788
789
790
791
792
793
794
795
                stream=True,
            )

            r.raise_for_status()

            return StreamingResponse(
                stream_content(),
                status_code=r.status_code,
                headers=dict(r.headers),
            )
        except Exception as e:
Timothy J. Baek's avatar
Timothy J. Baek committed
796
            print(e)
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
            raise e

    try:
        return await run_in_threadpool(get_request)
    except Exception as e:
        error_detail = "Open WebUI: Server Connection Error"
        if r is not None:
            try:
                res = r.json()
                if "error" in res:
                    error_detail = f"Ollama: {res['error']}"
            except:
                error_detail = f"Ollama: {e}"

        raise HTTPException(
            status_code=r.status_code if r else 500,
            detail=error_detail,
Timothy J. Baek's avatar
Timothy J. Baek committed
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
        )


# TODO: we should update this part once Ollama supports other types
class OpenAIChatMessage(BaseModel):
    role: str
    content: str

    model_config = ConfigDict(extra="allow")


class OpenAIChatCompletionForm(BaseModel):
    model: str
    messages: List[OpenAIChatMessage]

    model_config = ConfigDict(extra="allow")


@app.post("/v1/chat/completions")
@app.post("/v1/chat/completions/{url_idx}")
async def generate_openai_chat_completion(
    form_data: OpenAIChatCompletionForm,
    url_idx: Optional[int] = None,
    user=Depends(get_current_user),
):

    if url_idx == None:
        if form_data.model in app.state.MODELS:
            url_idx = random.choice(app.state.MODELS[form_data.model]["urls"])
        else:
            raise HTTPException(
                status_code=400,
                detail=ERROR_MESSAGES.MODEL_NOT_FOUND(form_data.model),
            )

    url = app.state.OLLAMA_BASE_URLS[url_idx]
850
    print(url)
Timothy J. Baek's avatar
Timothy J. Baek committed
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

    r = None

    def get_request():
        nonlocal form_data
        nonlocal r

        request_id = str(uuid.uuid4())
        try:
            REQUEST_POOL.append(request_id)

            def stream_content():
                try:
                    if form_data.stream:
                        yield json.dumps(
                            {"request_id": request_id, "done": False}
                        ) + "\n"

                    for chunk in r.iter_content(chunk_size=8192):
                        if request_id in REQUEST_POOL:
                            yield chunk
                        else:
                            print("User: canceled request")
                            break
                finally:
                    if hasattr(r, "close"):
                        r.close()
                        if request_id in REQUEST_POOL:
                            REQUEST_POOL.remove(request_id)

            r = requests.request(
                method="POST",
                url=f"{url}/v1/chat/completions",
Timothy J. Baek's avatar
Timothy J. Baek committed
884
                data=form_data.model_dump_json(exclude_none=True).encode(),
Timothy J. Baek's avatar
Timothy J. Baek committed
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
                stream=True,
            )

            r.raise_for_status()

            return StreamingResponse(
                stream_content(),
                status_code=r.status_code,
                headers=dict(r.headers),
            )
        except Exception as e:
            raise e

    try:
        return await run_in_threadpool(get_request)
    except Exception as e:
        error_detail = "Open WebUI: Server Connection Error"
        if r is not None:
            try:
                res = r.json()
                if "error" in res:
                    error_detail = f"Ollama: {res['error']}"
            except:
                error_detail = f"Ollama: {e}"

        raise HTTPException(
            status_code=r.status_code if r else 500,
            detail=error_detail,
913
914
915
        )


Timothy J. Baek's avatar
Timothy J. Baek committed
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
class UrlForm(BaseModel):
    url: str


class UploadBlobForm(BaseModel):
    filename: str


def parse_huggingface_url(hf_url):
    try:
        # Parse the URL
        parsed_url = urlparse(hf_url)

        # Get the path and split it into components
        path_components = parsed_url.path.split("/")

        # Extract the desired output
        user_repo = "/".join(path_components[1:3])
        model_file = path_components[-1]

        return model_file
    except ValueError:
        return None


async def download_file_stream(
    ollama_url, file_url, file_path, file_name, chunk_size=1024 * 1024
):
    done = False

    if os.path.exists(file_path):
        current_size = os.path.getsize(file_path)
    else:
        current_size = 0

    headers = {"Range": f"bytes={current_size}-"} if current_size > 0 else {}

    timeout = aiohttp.ClientTimeout(total=600)  # Set the timeout

    async with aiohttp.ClientSession(timeout=timeout) as session:
        async with session.get(file_url, headers=headers) as response:
            total_size = int(response.headers.get("content-length", 0)) + current_size

            with open(file_path, "ab+") as file:
                async for data in response.content.iter_chunked(chunk_size):
                    current_size += len(data)
                    file.write(data)

                    done = current_size == total_size
                    progress = round((current_size / total_size) * 100, 2)
Timothy J. Baek's avatar
Timothy J. Baek committed
966

Timothy J. Baek's avatar
Timothy J. Baek committed
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
                    yield f'data: {{"progress": {progress}, "completed": {current_size}, "total": {total_size}}}\n\n'

                if done:
                    file.seek(0)
                    hashed = calculate_sha256(file)
                    file.seek(0)

                    url = f"{ollama_url}/api/blobs/sha256:{hashed}"
                    response = requests.post(url, data=file)

                    if response.ok:
                        res = {
                            "done": done,
                            "blob": f"sha256:{hashed}",
                            "name": file_name,
                        }
                        os.remove(file_path)

                        yield f"data: {json.dumps(res)}\n\n"
                    else:
                        raise "Ollama: Could not create blob, Please try again."


Timothy J. Baek's avatar
Timothy J. Baek committed
990
991
992
993
994
# def number_generator():
#     for i in range(1, 101):
#         yield f"data: {i}\n"


Timothy J. Baek's avatar
Timothy J. Baek committed
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
# url = "https://huggingface.co/TheBloke/stablelm-zephyr-3b-GGUF/resolve/main/stablelm-zephyr-3b.Q2_K.gguf"
@app.post("/models/download")
@app.post("/models/download/{url_idx}")
async def download_model(
    form_data: UrlForm,
    url_idx: Optional[int] = None,
):

    if url_idx == None:
        url_idx = 0
    url = app.state.OLLAMA_BASE_URLS[url_idx]

    file_name = parse_huggingface_url(form_data.url)

    if file_name:
        file_path = f"{UPLOAD_DIR}/{file_name}"
        return StreamingResponse(
Timothy J. Baek's avatar
Timothy J. Baek committed
1012
            download_file_stream(url, form_data.url, file_path, file_name),
Timothy J. Baek's avatar
Timothy J. Baek committed
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
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
        )
    else:
        return None


@app.post("/models/upload")
@app.post("/models/upload/{url_idx}")
def upload_model(file: UploadFile = File(...), url_idx: Optional[int] = None):
    if url_idx == None:
        url_idx = 0
    ollama_url = app.state.OLLAMA_BASE_URLS[url_idx]

    file_path = f"{UPLOAD_DIR}/{file.filename}"

    # Save file in chunks
    with open(file_path, "wb+") as f:
        for chunk in file.file:
            f.write(chunk)

    def file_process_stream():
        nonlocal ollama_url
        total_size = os.path.getsize(file_path)
        chunk_size = 1024 * 1024
        try:
            with open(file_path, "rb") as f:
                total = 0
                done = False

                while not done:
                    chunk = f.read(chunk_size)
                    if not chunk:
                        done = True
                        continue

                    total += len(chunk)
                    progress = round((total / total_size) * 100, 2)

                    res = {
                        "progress": progress,
                        "total": total_size,
                        "completed": total,
                    }
                    yield f"data: {json.dumps(res)}\n\n"

                if done:
                    f.seek(0)
                    hashed = calculate_sha256(f)
                    f.seek(0)

Timothy J. Baek's avatar
Timothy J. Baek committed
1062
                    url = f"{ollama_url}/api/blobs/sha256:{hashed}"
Timothy J. Baek's avatar
Timothy J. Baek committed
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
                    response = requests.post(url, data=f)

                    if response.ok:
                        res = {
                            "done": done,
                            "blob": f"sha256:{hashed}",
                            "name": file.filename,
                        }
                        os.remove(file_path)
                        yield f"data: {json.dumps(res)}\n\n"
                    else:
                        raise Exception(
                            "Ollama: Could not create blob, Please try again."
                        )

        except Exception as e:
            res = {"error": str(e)}
            yield f"data: {json.dumps(res)}\n\n"

    return StreamingResponse(file_process_stream(), media_type="text/event-stream")


Timothy J. Baek's avatar
Timothy J. Baek committed
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
# async def upload_model(file: UploadFile = File(), url_idx: Optional[int] = None):
#     if url_idx == None:
#         url_idx = 0
#     url = app.state.OLLAMA_BASE_URLS[url_idx]

#     file_location = os.path.join(UPLOAD_DIR, file.filename)
#     total_size = file.size

#     async def file_upload_generator(file):
#         print(file)
#         try:
#             async with aiofiles.open(file_location, "wb") as f:
#                 completed_size = 0
#                 while True:
#                     chunk = await file.read(1024*1024)
#                     if not chunk:
#                         break
#                     await f.write(chunk)
#                     completed_size += len(chunk)
#                     progress = (completed_size / total_size) * 100

#                     print(progress)
#                     yield f'data: {json.dumps({"status": "uploading", "percentage": progress, "total": total_size, "completed": completed_size, "done": False})}\n'
#         except Exception as e:
#             print(e)
#             yield f"data: {json.dumps({'status': 'error', 'message': str(e)})}\n"
#         finally:
#             await file.close()
#             print("done")
#             yield f'data: {json.dumps({"status": "completed", "percentage": 100, "total": total_size, "completed": completed_size, "done": True})}\n'

#     return StreamingResponse(
#         file_upload_generator(copy.deepcopy(file)), media_type="text/event-stream"
#     )


1121
@app.api_route("/{path:path}", methods=["GET", "POST", "PUT", "DELETE"])
Timothy J. Baek's avatar
Timothy J. Baek committed
1122
async def deprecated_proxy(path: str, request: Request, user=Depends(get_current_user)):
1123
1124
    url = app.state.OLLAMA_BASE_URLS[0]
    target_url = f"{url}/{path}"
1125
1126
1127
1128
1129
1130
1131

    body = await request.body()
    headers = dict(request.headers)

    if user.role in ["user", "admin"]:
        if path in ["pull", "delete", "push", "copy", "create"]:
            if user.role != "admin":
Timothy J. Baek's avatar
Timothy J. Baek committed
1132
                raise HTTPException(
Timothy J. Baek's avatar
Timothy J. Baek committed
1133
1134
                    status_code=status.HTTP_401_UNAUTHORIZED,
                    detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
Timothy J. Baek's avatar
Timothy J. Baek committed
1135
                )
1136
    else:
Timothy J. Baek's avatar
Timothy J. Baek committed
1137
1138
1139
1140
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED,
            detail=ERROR_MESSAGES.ACCESS_PROHIBITED,
        )
1141

1142
1143
1144
1145
    headers.pop("host", None)
    headers.pop("authorization", None)
    headers.pop("origin", None)
    headers.pop("referer", None)
Timothy J. Baek's avatar
Timothy J. Baek committed
1146

Timothy J. Baek's avatar
Timothy J. Baek committed
1147
1148
1149
1150
    r = None

    def get_request():
        nonlocal r
1151
1152

        request_id = str(uuid.uuid4())
Timothy J. Baek's avatar
Timothy J. Baek committed
1153
        try:
1154
1155
1156
1157
            REQUEST_POOL.append(request_id)

            def stream_content():
                try:
1158
1159
1160
1161
1162
1163
1164
                    if path == "generate":
                        data = json.loads(body.decode("utf-8"))

                        if not ("stream" in data and data["stream"] == False):
                            yield json.dumps({"id": request_id, "done": False}) + "\n"

                    elif path == "chat":
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
                        yield json.dumps({"id": request_id, "done": False}) + "\n"

                    for chunk in r.iter_content(chunk_size=8192):
                        if request_id in REQUEST_POOL:
                            yield chunk
                        else:
                            print("User: canceled request")
                            break
                finally:
                    if hasattr(r, "close"):
                        r.close()
1176
1177
                        if request_id in REQUEST_POOL:
                            REQUEST_POOL.remove(request_id)
1178

Timothy J. Baek's avatar
Timothy J. Baek committed
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
            r = requests.request(
                method=request.method,
                url=target_url,
                data=body,
                headers=headers,
                stream=True,
            )

            r.raise_for_status()

1189
1190
            # r.close()

Timothy J. Baek's avatar
Timothy J. Baek committed
1191
            return StreamingResponse(
1192
                stream_content(),
Timothy J. Baek's avatar
Timothy J. Baek committed
1193
1194
1195
1196
1197
                status_code=r.status_code,
                headers=dict(r.headers),
            )
        except Exception as e:
            raise e
1198

Timothy J. Baek's avatar
Timothy J. Baek committed
1199
1200
    try:
        return await run_in_threadpool(get_request)
1201
    except Exception as e:
Timothy J. Baek's avatar
Timothy J. Baek committed
1202
        error_detail = "Open WebUI: Server Connection Error"
Timothy J. Baek's avatar
Timothy J. Baek committed
1203
        if r is not None:
1204
            try:
Timothy J. Baek's avatar
Timothy J. Baek committed
1205
                res = r.json()
1206
1207
1208
1209
1210
                if "error" in res:
                    error_detail = f"Ollama: {res['error']}"
            except:
                error_detail = f"Ollama: {e}"

Timothy J. Baek's avatar
Timothy J. Baek committed
1211
        raise HTTPException(
Timothy J. Baek's avatar
Timothy J. Baek committed
1212
            status_code=r.status_code if r else 500,
Timothy J. Baek's avatar
Timothy J. Baek committed
1213
1214
            detail=error_detail,
        )