gpt4free

公开

关注 0 Fork 0 Star 0

返回提交历史

XFEstudio/gpt4free

Refactor MarkItDown and OpenaiChat classes for improved media handling and optional parameters; enhance is_data_an_media function to support binary/octet-stream return type for unsupported URLs.

2317bd5a

hlohaus <983577+hlohaus@users.noreply.github.com>

提交于 2025-10-31 18:16

父提交 35e3fa9

代码差异

4 个文件 +19 -25

Modified g4f/Provider/audio/MarkItDown.py +5 -13

@@ -2,7 +2,6 @@ from __future__ import annotations

2 2

3 3 import os

4 4 import asyncio

5 from typing import Any

6 5

7 6 try:

8 7 from ...integration.markitdown import MarkItDown as MaItDo, StreamInfo

@@ -23,7 +22,6 @@ class MarkItDown(AsyncGeneratorProvider, ProviderModelMixin):

23 22 model: str,

24 23 messages: Messages,

25 24 media: MediaListType = None,

26 llm_client: Any = None,

27 25 **kwargs

28 26 ) -> AsyncResult:

29 27 if media is None:

@@ -34,12 +32,10 @@ class MarkItDown(AsyncGeneratorProvider, ProviderModelMixin):

34 32 for file, filename in media:

35 33 text = None

36 34 try:

37 result = md.convert(

38 file,

39 stream_info=StreamInfo(filename=filename) if filename else None,

40 llm_client=llm_client,

41 llm_model=model

42 )

35 if isinstance(file, str) and file.startswith(("http://", "https://")):

36 result = md.convert_url(file)

37 else:

38 result = md.convert(file, stream_info=StreamInfo(filename=filename) if filename else None)

43 39 if asyncio.iscoroutine(result.text_content):

44 40 text = await result.text_content

45 41 else:

@@ -47,11 +43,7 @@ class MarkItDown(AsyncGeneratorProvider, ProviderModelMixin):

47 43 except TypeError:

48 44 copyfile = get_tempfile(file, filename)

49 45 try:

50 result = md.convert(

51 copyfile,

52 llm_client=llm_client,

53 llm_model=model

54 )

46 result = md.convert(copyfile)

55 47 if asyncio.iscoroutine(result.text_content):

56 48 text = await result.text_content

57 49 else:

Modified g4f/Provider/needs_auth/OpenaiChat.py +11 -7

@@ -24,7 +24,7 @@ from ...requests import StreamSession

24 24 from ...requests import get_nodriver_session

25 25 from ...image import ImageRequest, to_image, to_bytes, is_accepted_format, detect_file_type

26 26 from ...errors import MissingAuthError, NoValidHarFileError, ModelNotFoundError

from ...providers.response import JsonConversation, FinishReason, SynthesizeData, AuthResult, ImageResponse, ImagePreview, ResponseType, format_link

from ...providers.response import JsonConversation, FinishReason, SynthesizeData, AuthResult, ImageResponse, ImagePreview, ResponseType, JsonRequest, format_link

28 28 from ...providers.response import TitleGeneration, RequestLogin, Reasoning

29 29 from ...tools.media import merge_media

30 30 from ..helper import format_cookies, format_media_prompt, to_string

@@ -330,14 +330,15 @@ class OpenaiChat(AsyncAuthedProvider, ProviderModelMixin):

330 330 proxy: str = None,

331 331 timeout: int = 360,

332 332 auto_continue: bool = False,

333 action: str = "next",

333 action: Optional[str] = None,

334 334 conversation: Conversation = None,

335 335 media: MediaListType = None,

336 336 return_conversation: bool = True,

337 337 web_search: bool = False,

338 338 prompt: str = None,

339 conversation_mode=None,

340 temporary=False,

339 conversation_mode: Optional[dict] = None,

340 temporary: Optional[bool] = None,

341 conversation_id: Optional[str] = None,

341 342 **kwargs

342 343 ) -> AsyncResult:

343 344 """

@@ -351,7 +352,6 @@ class OpenaiChat(AsyncAuthedProvider, ProviderModelMixin):

351 352 api_key (str): Access token for authentication.

352 353 auto_continue (bool): Flag to automatically continue the conversation.

353 354 action (str): Type of action ('next', 'continue', 'variant').

354 conversation_id (str): ID of the conversation.

355 355 media (MediaListType): Images to include in the conversation.

356 356 return_conversation (bool): Flag to include response fields in the output.

357 357 **kwargs: Additional keyword arguments.

@@ -362,6 +362,10 @@ class OpenaiChat(AsyncAuthedProvider, ProviderModelMixin):

362 362 Raises:

363 363 RuntimeError: If an error occurs during processing.

364 364 """

365 if temporary is None:

366 temporary = action is not None and conversation_id is not None

367 if action is None:

368 action = "next"

365 369 async with StreamSession(

366 370 proxy=proxy,

367 371 impersonate="chrome",

@@ -431,7 +435,6 @@ class OpenaiChat(AsyncAuthedProvider, ProviderModelMixin):

431 435 }

432 436 if temporary:

433 437 data["history_and_training_disabled"] = True

434

435 438 async with session.post(

436 439 prepare_url,

437 440 json=data,

@@ -494,7 +497,7 @@ class OpenaiChat(AsyncAuthedProvider, ProviderModelMixin):

494 497 if temporary:

495 498 data["history_and_training_disabled"] = True

496 499

497 if conversation.conversation_id is not None:

500 if conversation.conversation_id is not None and not temporary:

498 501 data["conversation_id"] = conversation.conversation_id

499 502 debug.log(f"OpenaiChat: Use conversation: {conversation.conversation_id}")

500 503 prompt = conversation.prompt = format_media_prompt(messages, prompt)

@@ -510,6 +513,7 @@ class OpenaiChat(AsyncAuthedProvider, ProviderModelMixin):

510 513 else:

511 514 new_messages.append(message)

512 515

                    data["messages"] = cls.create_messages(new_messages, image_requests, ["search"] if web_search else None)

516 yield JsonRequest.from_dict(data)

513 517 headers = {

514 518 **cls._headers,

515 519 "accept": "text/event-stream",

Modified g4f/image/__init__.py +3 -2

@@ -107,11 +107,12 @@ def is_data_an_media(data, filename: str = None) -> str:

107 107 return content_type

108 108 if isinstance(data, bytes):

109 109 return is_accepted_format(data)

110 if isinstance(data, str) and data.startswith("http"):

110 if isinstance(data, str) and data.startswith(("http://", "https://")):

111 111 path = urlparse(data).path

112 112 extension = get_extension(path)

113 113 if extension is not None:

114 return EXTENSIONS_MAP[extension]

115 return "binary/octet-stream"

115 116 return is_data_uri_an_image(data)

116 117

117 118 def is_valid_media(data: ImageType = None, filename: str = None) -> str:

Modified g4f/integration/markitdown/_youtube_converter.py +0 -3

@@ -75,7 +75,6 @@ class YouTubeConverter(DocumentConverter):

75 75 ) -> DocumentConverterResult:

76 76 # Parse the stream

77 77 encoding = "utf-8" if stream_info.charset is None else stream_info.charset

78 print(file_stream)

79 78 soup = bs4.BeautifulSoup(file_stream, "html.parser", from_encoding=encoding)

80 79

81 80 # Read the meta tags

@@ -95,8 +94,6 @@ class YouTubeConverter(DocumentConverter):

95 94 if key and content: # Only add non-empty content

96 95 metadata[key] = content

97 96 break

99 print(f"Extracted metadata keys: {list(metadata.keys())}")

100 97

101 98 # Try reading the description

102 99 try: