summaryrefslogtreecommitdiffstats
path: root/g4f/Provider/needs_auth/OpenaiChat.py
blob: 8c2668aba17c53a2d4e6d4161d84313935649638 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
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
201
202
203
204
205
206
207
208
209
210
211
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
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
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
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
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
380
381
382
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
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
449
450
451
452
453
454
455
456
457
458
459
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
496
497
498
499
500
501
502
503
504
505
506
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
535
536
537
538
539
540
541
542
543
544
545
546
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
588
589
590
591
592
593
594
595
596
597
598
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
from __future__ import annotations

import asyncio
import uuid
import json
import os

try:
    from py_arkose_generator.arkose import get_values_for_request
    from async_property import async_cached_property
    has_requirements = True
except ImportError:
    async_cached_property = property
    has_requirements = False
try:
    from selenium.webdriver.common.by import By
    from selenium.webdriver.support.ui import WebDriverWait
    from selenium.webdriver.support import expected_conditions as EC
except ImportError:
    pass

from ..base_provider import AsyncGeneratorProvider, ProviderModelMixin
from ..helper import get_cookies
from ...webdriver import get_browser
from ...typing import AsyncResult, Messages, Cookies, ImageType, Union
from ...requests import get_args_from_browser
from ...requests.aiohttp import StreamSession
from ...image import to_image, to_bytes, ImageResponse, ImageRequest
from ...errors import MissingRequirementsError, MissingAuthError
from ... import debug

class OpenaiChat(AsyncGeneratorProvider, ProviderModelMixin):
    """A class for creating and managing conversations with OpenAI chat service"""
    
    url = "https://chat.openai.com"
    working = True
    needs_auth = True
    supports_gpt_35_turbo = True
    supports_gpt_4 = True
    supports_message_history = True
    default_model = None
    models = ["gpt-3.5-turbo", "gpt-4", "gpt-4-gizmo"]
    model_aliases = {"text-davinci-002-render-sha": "gpt-3.5-turbo"}
    _args: dict = None

    @classmethod
    async def create(
        cls,
        prompt: str = None,
        model: str = "",
        messages: Messages = [],
        history_disabled: bool = False,
        action: str = "next",
        conversation_id: str = None,
        parent_id: str = None,
        image: ImageType = None,
        **kwargs
    ) -> Response:
        """
        Create a new conversation or continue an existing one
        
        Args:
            prompt: The user input to start or continue the conversation
            model: The name of the model to use for generating responses
            messages: The list of previous messages in the conversation
            history_disabled: A flag indicating if the history and training should be disabled
            action: The type of action to perform, either "next", "continue", or "variant"
            conversation_id: The ID of the existing conversation, if any
            parent_id: The ID of the parent message, if any
            image: The image to include in the user input, if any
            **kwargs: Additional keyword arguments to pass to the generator
        
        Returns:
            A Response object that contains the generator, action, messages, and options
        """
        # Add the user input to the messages list
        if prompt:
            messages.append({
                "role": "user",
                "content": prompt
            })
        generator = cls.create_async_generator(
            model,
            messages,
            history_disabled=history_disabled,
            action=action,
            conversation_id=conversation_id,
            parent_id=parent_id,
            image=image,
            response_fields=True,
            **kwargs
        )
        return Response(
            generator,
            action,
            messages,
            kwargs
        )
    
    @classmethod
    async def upload_image(
        cls,
        session: StreamSession,
        headers: dict,
        image: ImageType,
        image_name: str = None
    ) -> ImageRequest:
        """
        Upload an image to the service and get the download URL
        
        Args:
            session: The StreamSession object to use for requests
            headers: The headers to include in the requests
            image: The image to upload, either a PIL Image object or a bytes object
        
        Returns:
            An ImageRequest object that contains the download URL, file name, and other data
        """
        # Convert the image to a PIL Image object and get the extension
        image = to_image(image)
        extension = image.format.lower()
        # Convert the image to a bytes object and get the size
        data_bytes = to_bytes(image)
        data = {
            "file_name": image_name if image_name else f"{image.width}x{image.height}.{extension}",
            "file_size": len(data_bytes),
            "use_case":	"multimodal"
        }
        # Post the image data to the service and get the image data
        async with session.post(f"{cls.url}/backend-api/files", json=data, headers=headers) as response:
            response.raise_for_status()
            image_data = {
                **data,
                **await response.json(),
                "mime_type": f"image/{extension}",
                "extension": extension,
                "height": image.height,
                "width": image.width
            }
        # Put the image bytes to the upload URL and check the status
        async with session.put(
            image_data["upload_url"],
            data=data_bytes,
            headers={
                "Content-Type": image_data["mime_type"],
                "x-ms-blob-type": "BlockBlob"
            }
        ) as response:
            response.raise_for_status()
        # Post the file ID to the service and get the download URL
        async with session.post(
            f"{cls.url}/backend-api/files/{image_data['file_id']}/uploaded",
            json={},
            headers=headers
        ) as response:
            response.raise_for_status()
            image_data["download_url"] = (await response.json())["download_url"]
        return ImageRequest(image_data)
    
    @classmethod
    async def get_default_model(cls, session: StreamSession, headers: dict):
        """
        Get the default model name from the service
        
        Args:
            session: The StreamSession object to use for requests
            headers: The headers to include in the requests
        
        Returns:
            The default model name as a string
        """
        if not cls.default_model:
            async with session.get(f"{cls.url}/backend-api/models", headers=headers) as response:
                cls._update_request_args(session)
                response.raise_for_status()
                data = await response.json()
                if "categories" in data:
                    cls.default_model = data["categories"][-1]["default_model"]
                    return cls.default_model 
                raise RuntimeError(f"Response: {data}")
        return cls.default_model
    
    @classmethod
    def create_messages(cls, messages: Messages, image_request: ImageRequest = None):
        """
        Create a list of messages for the user input
        
        Args:
            prompt: The user input as a string
            image_response: The image response object, if any
        
        Returns:
            A list of messages with the user input and the image, if any
        """
        # Create a message object with the user role and the content
        messages = [{
            "id": str(uuid.uuid4()),
            "author": {"role": message["role"]},
            "content": {"content_type": "text", "parts": [message["content"]]},
        } for message in messages]

        # Check if there is an image response
        if image_request:
            # Change content in last user message
            messages[-1]["content"] = {
                "content_type": "multimodal_text",
                "parts": [{
                    "asset_pointer": f"file-service://{image_request.get('file_id')}",
                    "height": image_request.get("height"),
                    "size_bytes": image_request.get("file_size"),
                    "width": image_request.get("width"),
                }, messages[-1]["content"]["parts"][0]]
            }
            # Add the metadata object with the attachments
            messages[-1]["metadata"] = {
                "attachments": [{
                    "height": image_request.get("height"),
                    "id": image_request.get("file_id"),
                    "mimeType": image_request.get("mime_type"),
                    "name": image_request.get("file_name"),
                    "size": image_request.get("file_size"),
                    "width": image_request.get("width"),
                }]
            }
        return messages

    @classmethod
    async def get_generated_image(cls, session: StreamSession, headers: dict, line: dict) -> ImageResponse:
        """
        Retrieves the image response based on the message content.

        This method processes the message content to extract image information and retrieves the 
        corresponding image from the backend API. It then returns an ImageResponse object containing 
        the image URL and the prompt used to generate the image.

        Args:
            session (StreamSession): The StreamSession object used for making HTTP requests.
            headers (dict): HTTP headers to be used for the request.
            line (dict): A dictionary representing the line of response that contains image information.

        Returns:
            ImageResponse: An object containing the image URL and the prompt, or None if no image is found.

        Raises:
            RuntimeError: If there'san error in downloading the image, including issues with the HTTP request or response.
        """
        if "parts" not in line["message"]["content"]:
            return
        first_part = line["message"]["content"]["parts"][0]
        if "asset_pointer" not in first_part or "metadata" not in first_part:
            return
        if first_part["metadata"] is None:
            return
        prompt = first_part["metadata"]["dalle"]["prompt"]
        file_id = first_part["asset_pointer"].split("file-service://", 1)[1]
        try:
            async with session.get(f"{cls.url}/backend-api/files/{file_id}/download", headers=headers) as response:
                response.raise_for_status()
                download_url = (await response.json())["download_url"]
                return ImageResponse(download_url, prompt)
        except Exception as e:
            raise RuntimeError(f"Error in downloading image: {e}")

    @classmethod
    async def delete_conversation(cls, session: StreamSession, headers: dict, conversation_id: str):
        """
        Deletes a conversation by setting its visibility to False.

        This method sends an HTTP PATCH request to update the visibility of a conversation. 
        It's used to effectively delete a conversation from being accessed or displayed in the future.

        Args:
            session (StreamSession): The StreamSession object used for making HTTP requests.
            headers (dict): HTTP headers to be used for the request.
            conversation_id (str): The unique identifier of the conversation to be deleted.

        Raises:
            HTTPError: If the HTTP request fails or returns an unsuccessful status code.
        """
        async with session.patch(
            f"{cls.url}/backend-api/conversation/{conversation_id}",
            json={"is_visible": False},
            headers=headers
        ) as response:
            response.raise_for_status()

    @classmethod
    async def create_async_generator(
        cls,
        model: str,
        messages: Messages,
        proxy: str = None,
        timeout: int = 120,
        api_key: str = None,
        cookies: Cookies = None,
        auto_continue: bool = False,
        history_disabled: bool = True,
        action: str = "next",
        conversation_id: str = None,
        parent_id: str = None,
        image: ImageType = None,
        response_fields: bool = False,
        **kwargs
    ) -> AsyncResult:
        """
        Create an asynchronous generator for the conversation.

        Args:
            model (str): The model name.
            messages (Messages): The list of previous messages.
            proxy (str): Proxy to use for requests.
            timeout (int): Timeout for requests.
            api_key (str): Access token for authentication.
            cookies (dict): Cookies to use for authentication.
            auto_continue (bool): Flag to automatically continue the conversation.
            history_disabled (bool): Flag to disable history and training.
            action (str): Type of action ('next', 'continue', 'variant').
            conversation_id (str): ID of the conversation.
            parent_id (str): ID of the parent message.
            image (ImageType): Image to include in the conversation.
            response_fields (bool): Flag to include response fields in the output.
            **kwargs: Additional keyword arguments.

        Yields:
            AsyncResult: Asynchronous results from the generator.

        Raises:
            RuntimeError: If an error occurs during processing.
        """
        if not has_requirements:
            raise MissingRequirementsError('Install "py-arkose-generator" and "async_property" package')
        if not parent_id:
            parent_id = str(uuid.uuid4())

        # Read api_key from args
        api_key = kwargs["access_token"] if "access_token" in kwargs else api_key
        if cls._args is None:
            if api_key is None:
                # Read api_key from cookies
                cookies = get_cookies("chat.openai.com", False) if cookies is None else cookies
                api_key = cookies["access_token"] if "access_token" in cookies else api_key
            cls._args = cls._create_request_args(cookies)
 
        async with StreamSession(
            proxies={"https": proxy},
            impersonate="chrome",
            timeout=timeout
        ) as session:
            if api_key is None and cookies:
                # Read api_key from session
                api_key = await cls.fetch_access_token(session, cls._args["headers"])

            if api_key is not None:
                cls._args["headers"]["Authorization"] = f"Bearer {api_key}"
                try:
                    cls.default_model = await cls.get_default_model(session, cls._args["headers"])
                except Exception as e:
                    if debug.logging:
                        print(f"{e.__class__.__name__}: {e}")

            if cls.default_model is None:
                login_url = os.environ.get("G4F_LOGIN_URL")
                if login_url:
                    yield f"Please login: [ChatGPT]({login_url})\n\n"
                try:
                    cls._args = cls.browse_access_token(proxy)
                except MissingRequirementsError:
                    raise MissingAuthError(f'Missing or invalid "access_token". Add a new "api_key" please')
                cls.default_model = await cls.get_default_model(session, cls._args["headers"])

            try:
                image_response = await cls.upload_image(
                    session,
                    cls._args["headers"],
                    image,
                    kwargs.get("image_name")
                ) if image else None
            except Exception as e:
                yield e

            end_turn = EndTurn()
            model = cls.get_model(model)
            model = "text-davinci-002-render-sha" if model == "gpt-3.5-turbo" else model
            while not end_turn.is_end:
                arkose_token = await cls.get_arkose_token(session)
                data = {
                    "action": action,
                    "arkose_token": arkose_token,
                    "conversation_mode": {"kind": "primary_assistant"},
                    "force_paragen": False,
                    "force_rate_limit": False,
                    "conversation_id": conversation_id,
                    "parent_message_id": parent_id,
                    "model": model,
                    "history_and_training_disabled": history_disabled and not auto_continue,
                }
                if action != "continue":
                    messages = messages if not conversation_id else [messages[-1]]
                    data["messages"] = cls.create_messages(messages, image_response)                

                async with session.post(
                    f"{cls.url}/backend-api/conversation",
                    json=data,
                    headers={
                        "Accept": "text/event-stream",
                        "OpenAI-Sentinel-Arkose-Token": arkose_token,
                        **cls._args["headers"]
                    }
                ) as response:
                    cls._update_request_args(session)
                    if not response.ok:
                        raise RuntimeError(f"Response {response.status}: {await response.text()}")
                    last_message: int = 0
                    async for line in response.iter_lines():
                        if not line.startswith(b"data: "):
                            continue
                        elif line.startswith(b"data: [DONE]"):
                            break
                        try:
                            line = json.loads(line[6:])
                        except:
                            continue
                        if "message" not in line:
                            continue
                        if "error" in line and line["error"]:
                            raise RuntimeError(line["error"])
                        if "message_type" not in line["message"]["metadata"]:
                            continue
                        try:
                            image_response = await cls.get_generated_image(session, cls._args["headers"], line)
                            if image_response is not None:
                                yield image_response
                        except Exception as e:
                            yield e
                        if line["message"]["author"]["role"] != "assistant":
                            continue
                        if line["message"]["content"]["content_type"] != "text":
                            continue
                        if line["message"]["metadata"]["message_type"] not in ("next", "continue", "variant"):
                            continue
                        conversation_id = line["conversation_id"]
                        parent_id = line["message"]["id"]
                        if response_fields:
                            response_fields = False
                            yield ResponseFields(conversation_id, parent_id, end_turn)
                        if "parts" in line["message"]["content"]:
                            new_message = line["message"]["content"]["parts"][0]
                            if len(new_message) > last_message:
                                yield new_message[last_message:]
                            last_message = len(new_message)
                        if "finish_details" in line["message"]["metadata"]:
                            if line["message"]["metadata"]["finish_details"]["type"] == "stop":
                                end_turn.end()
                if not auto_continue:
                    break
                action = "continue"
                await asyncio.sleep(5)
            if history_disabled and auto_continue:
                await cls.delete_conversation(session, cls._args["headers"], conversation_id)

    @classmethod
    def browse_access_token(cls, proxy: str = None, timeout: int = 1200) -> tuple[str, dict]:
        """
        Browse to obtain an access token.

        Args:
            proxy (str): Proxy to use for browsing.

        Returns:
            tuple[str, dict]: A tuple containing the access token and cookies.
        """
        driver = get_browser(proxy=proxy)
        try:
            driver.get(f"{cls.url}/")
            WebDriverWait(driver, timeout).until(EC.presence_of_element_located((By.ID, "prompt-textarea")))
            access_token = driver.execute_script(
                "let session = await fetch('/api/auth/session');"
                "let data = await session.json();"
                "let accessToken = data['accessToken'];"
                "let expires = new Date(); expires.setTime(expires.getTime() + 60 * 60 * 4 * 1000);"
                "document.cookie = 'access_token=' + accessToken + ';expires=' + expires.toUTCString() + ';path=/';"
                "return accessToken;"
            )
            args = get_args_from_browser(f"{cls.url}/", driver, do_bypass_cloudflare=False)
            args["headers"]["Authorization"] = f"Bearer {access_token}"
            args["headers"]["Cookie"] = cls._format_cookies(args["cookies"])
            return args
        finally:
            driver.close() 

    @classmethod
    async def get_arkose_token(cls, session: StreamSession) -> str:
        """
        Obtain an Arkose token for the session.

        Args:
            session (StreamSession): The session object.

        Returns:
            str: The Arkose token.

        Raises:
            RuntimeError: If unable to retrieve the token.
        """
        config = {
            "pkey": "3D86FBBA-9D22-402A-B512-3420086BA6CC",
            "surl": "https://tcr9i.chat.openai.com",
            "headers": {
                "User-Agent": 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/107.0.0.0 Safari/537.36'
            },
            "site": cls.url,
        }
        args_for_request = get_values_for_request(config)
        async with session.post(**args_for_request) as response:
            response.raise_for_status()
            decoded_json = await response.json()
            if "token" in decoded_json:
                return decoded_json["token"]
            raise RuntimeError(f"Response: {decoded_json}")

    @classmethod
    async def fetch_access_token(cls, session: StreamSession, headers: dict):
        async with session.get(
            f"{cls.url}/api/auth/session",
            headers=headers
        ) as response:
            if response.ok:
                data = await response.json()
                if "accessToken" in data:
                    return data["accessToken"]

    @staticmethod
    def _format_cookies(cookies: Cookies):
        return "; ".join(f"{k}={v}" for k, v in cookies.items() if k != "access_token")

    @classmethod
    def _create_request_args(cls, cookies: Union[Cookies, None]):
        return {
            "headers": {} if cookies is None else {"Cookie": cls._format_cookies(cookies)},
            "cookies": {} if cookies is None else cookies
        }

    @classmethod
    def _update_request_args(cls, session: StreamSession):
        for c in session.cookie_jar if hasattr(session, "cookie_jar") else session.cookies.jar:
            cls._args["cookies"][c.name if hasattr(c, "name") else c.key] = c.value
        cls._args["headers"]["Cookie"] = cls._format_cookies(cls._args["cookies"])

class EndTurn:
    """
    Class to represent the end of a conversation turn.
    """
    def __init__(self):
        self.is_end = False

    def end(self):
        self.is_end = True

class ResponseFields:
    """
    Class to encapsulate response fields.
    """
    def __init__(self, conversation_id: str, message_id: str, end_turn: EndTurn):
        self.conversation_id = conversation_id
        self.message_id = message_id
        self._end_turn = end_turn

class Response():
    """
    Class to encapsulate a response from the chat service.
    """
    def __init__(
        self,
        generator: AsyncResult,
        action: str,
        messages: Messages,
        options: dict
    ):
        self._generator = generator
        self.action = action
        self.is_end = False
        self._message = None
        self._messages = messages
        self._options = options
        self._fields = None

    async def generator(self):
        if self._generator:
            self._generator = None
            chunks = []
            async for chunk in self._generator:
                if isinstance(chunk, ResponseFields):
                    self._fields = chunk
                else:
                    yield chunk
                    chunks.append(str(chunk))
            self._message = "".join(chunks)
            if not self._fields:
                raise RuntimeError("Missing response fields")
            self.is_end = self._fields._end_turn.is_end

    def __aiter__(self):
        return self.generator()

    @async_cached_property
    async def message(self) -> str:
        await self.generator()
        return self._message

    async def get_fields(self):
        await self.generator()
        return {"conversation_id": self._fields.conversation_id, "parent_id": self._fields.message_id}

    async def next(self, prompt: str, **kwargs) -> Response:
        return await OpenaiChat.create(
            **self._options,
            prompt=prompt,
            messages=await self.messages,
            action="next",
            **await self.get_fields(),
            **kwargs
        )

    async def do_continue(self, **kwargs) -> Response:
        fields = await self.get_fields()
        if self.is_end:
            raise RuntimeError("Can't continue message. Message already finished.")
        return await OpenaiChat.create(
            **self._options,
            messages=await self.messages,
            action="continue",
            **fields,
            **kwargs
        )

    async def variant(self, **kwargs) -> Response:
        if self.action != "next":
            raise RuntimeError("Can't create variant from continue or variant request.")
        return await OpenaiChat.create(
            **self._options,
            messages=self._messages,
            action="variant",
            **await self.get_fields(),
            **kwargs
        )

    @async_cached_property
    async def messages(self):
        messages = self._messages
        messages.append({"role": "assistant", "content": await self.message})
        return messages