unpatch HuggingChat

@@ -1,20 +1,17 @@

1 1 from __future__ import annotations

2 2

3 import json

4 import requests

5 from aiohttp import ClientSession, BaseConnector

3 import json, requests, re

6 4

7 from ..typing import AsyncResult, Messages

8 from ..requests.raise_for_status import raise_for_status

9 from ..providers.conversation import BaseConversation

10 from .base_provider import AsyncGeneratorProvider, ProviderModelMixin

11 from .helper import format_prompt, get_connector, get_cookies

5 from curl_cffi import requests as cf_reqs

6 from ..typing import CreateResult, Messages

7 from .base_provider import ProviderModelMixin, AbstractProvider

8 from .helper import format_prompt, get_connector, get_cookies

12 9

13 class HuggingChat(AsyncGeneratorProvider, ProviderModelMixin):

14 url = "https://huggingface.co/chat"

15 working = True

16 needs_auth = True

17 default_model = "mistralai/Mixtral-8x7B-Instruct-v0.1"

10 class HuggingChat(AbstractProvider, ProviderModelMixin):

11 url = "https://huggingface.co/chat"

12 working = True

13 supports_stream = True

14 default_model = "mistralai/Mixtral-8x7B-Instruct-v0.1"

18 15 models = [

19 16 "HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1",

20 17 'CohereForAI/c4ai-command-r-plus',

@@ -26,92 +23,118 @@ class HuggingChat(AsyncGeneratorProvider, ProviderModelMixin):

26 23 'microsoft/Phi-3-mini-4k-instruct',

27 24 '01-ai/Yi-1.5-34B-Chat'

28 25 ]

26

29 27 model_aliases = {

30 28 "mistralai/Mistral-7B-Instruct-v0.1": "mistralai/Mistral-7B-Instruct-v0.2"

31 29 }

32 30

33 31 @classmethod

34 def get_models(cls):

35 if not cls.models:

36 url = f"{cls.url}/__data.json"

37 data = requests.get(url).json()["nodes"][0]["data"]

38 models = [data[key]["name"] for key in data[data[0]["models"]]]

39 cls.models = [data[key] for key in models]

40 return cls.models

41

42 @classmethod

43 async def create_async_generator(

32 def create_completion(

44 33 cls,

45 34 model: str,

46 35 messages: Messages,

47 stream: bool = True,

48 proxy: str = None,

49 connector: BaseConnector = None,

50 web_search: bool = False,

51 cookies: dict = None,

52 conversation: Conversation = None,

53 return_conversation: bool = False,

54 delete_conversation: bool = True,

36 stream: bool,

55 37 **kwargs

56 ) -> AsyncResult:

57 options = {"model": cls.get_model(model)}

58 if cookies is None:

59 cookies = get_cookies("huggingface.co", False)

60 if return_conversation:

61 delete_conversation = False

38 ) -> CreateResult:

39

40 if (model in cls.models) :

41

42 session = requests.Session()

43 headers = {

44 'accept': '*/*',

45 'accept-language': 'en,fr-FR;q=0.9,fr;q=0.8,es-ES;q=0.7,es;q=0.6,en-US;q=0.5,am;q=0.4,de;q=0.3',

46 'cache-control': 'no-cache',

47 'origin': 'https://huggingface.co',

48 'pragma': 'no-cache',

49 'priority': 'u=1, i',

50 'referer': 'https://huggingface.co/chat/',

51 'sec-ch-ua': '"Not/A)Brand";v="8", "Chromium";v="126", "Google Chrome";v="126"',

52 'sec-ch-ua-mobile': '?0',

53 'sec-ch-ua-platform': '"macOS"',

54 'sec-fetch-dest': 'empty',

55 'sec-fetch-mode': 'cors',

56 'sec-fetch-site': 'same-origin',

57

                'user-agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/126.0.0.0 Safari/537.36',

58 }

62 59

63 system_prompt = "\n".join([message["content"] for message in messages if message["role"] == "system"])

64 if system_prompt:

65 options["preprompt"] = system_prompt

66 messages = [message for message in messages if message["role"] != "system"]

67 headers = {

68

            'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/111.0.0.0 Safari/537.36',

69 }

70 async with ClientSession(

71 cookies=cookies,

72 headers=headers,

73 connector=get_connector(connector, proxy)

74 ) as session:

75 if conversation is None:

76 async with session.post(f"{cls.url}/conversation", json=options) as response:

77 await raise_for_status(response)

78 conversation_id = (await response.json())["conversationId"]

79 if return_conversation:

80 yield Conversation(conversation_id)

81 else:

82 conversation_id = conversation.conversation_id

83 async with session.get(f"{cls.url}/conversation/{conversation_id}/__data.json") as response:

84 await raise_for_status(response)

85 data: list = (await response.json())["nodes"][1]["data"]

86 keys: list[int] = data[data[0]["messages"]]

87 message_keys: dict = data[keys[0]]

88 message_id: str = data[message_keys["id"]]

89 options = {

90 "id": message_id,

91 "inputs": format_prompt(messages) if conversation is None else messages[-1]["content"],

92 "is_continue": False,

60 json_data = {

61 'searchEnabled': True,

62 'ethicsModalAccepted': True,

63 'ethicsModalAcceptedAt': None,

64 'activeModel': 'CohereForAI/c4ai-command-r-plus', # doesn't matter

65 'hideEmojiOnSidebar': False,

66 'shareConversationsWithModelAuthors': False,

67 'customPrompts': {},

68 'assistants': [],

69 'tools': {},

70 'disableStream': False,

71 'recentlySaved': False,

72 }

73

74 response = cf_reqs.post('https://huggingface.co/chat/settings', headers=headers, json=json_data)

75 session.cookies.update(response.cookies)

76

77 response = session.post('https://huggingface.co/chat/conversation',

78 headers=headers, json={'model': model})

79

80 conversationId = response.json()['conversationId']

81

            response       = session.get(f'https://huggingface.co/chat/conversation/{conversationId}/__data.json?x-sveltekit-invalidated=11',

82 headers=headers,

83 )

84

85 messageId = extract_id(response.json())

86

87 settings = {

88 "inputs": format_prompt(messages),

89 "id": messageId,

93 90 "is_retry": False,

94 "web_search": web_search

91 "is_continue": False,

92 "web_search": False,

93

94 # enable tools

95 "tools": {

96 "websearch": True,

97 "document_parser": False,

98 "query_calculator": False,

99 "image_generation": False,

100 "image_editing": False,

101 "fetch_url": False,

102 }

103 }

104

105 payload = {

106 "data": json.dumps(settings),

95 107 }

96 async with session.post(f"{cls.url}/conversation/{conversation_id}", json=options) as response:

97 first_token = True

98 async for line in response.content:

99 await raise_for_status(response)

100 line = json.loads(line)

101 if "type" not in line:

102 raise RuntimeError(f"Response: {line}")

103 elif line["type"] == "stream":

104 token = line["token"]

105 if first_token:

106 token = token.lstrip().replace('\u0000', '')

107 first_token = False

108 yield token

109 elif line["type"] == "finalAnswer":

110 break

111 if delete_conversation:

112 async with session.delete(f"{cls.url}/conversation/{conversation_id}") as response:

113 await raise_for_status(response)

114 108

115 class Conversation(BaseConversation):

116 def __init__(self, conversation_id: str) -> None:

117 self.conversation_id = conversation_id

109 response = session.post(f"https://huggingface.co/chat/conversation/{conversationId}",

110 headers=headers, data=payload, stream=True,

111 )

112

113 first_token = True

114 for line in response.iter_lines():

115 line = json.loads(line)

116

117 if "type" not in line:

118 raise RuntimeError(f"Response: {line}")

119

120 elif line["type"] == "stream":

121 token = line["token"]

122 if first_token:

123 token = token.lstrip().replace('\u0000', '')

124 first_token = False

125 #yield token

126 yield (token)

127

128 elif line["type"] == "finalAnswer":

129 break

130

131 def extract_id(response: dict) -> str:

132 data = response["nodes"][1]["data"]

133 uuid_pattern = re.compile(

134 r"^[\da-f]{8}-[\da-f]{4}-[\da-f]{4}-[\da-f]{4}-[\da-f]{12}$"

135 )

136 for item in data:

137 if type(item) == str and uuid_pattern.match(item):

138 return item

139

140 return None

gpt4free

代码差异