MeUtils 2025.3.26.18.50.26__py3-none-any.whl → 2025.4.1.18.47.48__py3-none-any.whl

meutils/apis/search/{web_search.py → zhipu_web_search.py}

@@ -6,10 +6,11 @@
 # @Author       : betterme
 # @WeChat       : meutils
 # @Software     : PyCharm
-# @Description  : 
+# @Description  :
 
-from aiostream import stream
 from meutils.pipe import *
+from meutils.caches import rcache
+
 from meutils.async_utils import sync_to_async
 
 from meutils.llm.clients import zhipuai_sdk_client
@@ -39,17 +40,14 @@ class Completions(object):
     def __init__(self, api_key: Optional[str] = None):
         self.api_key = api_key
 
-    @sync_to_async(thread_sensitive=False)
-    def create(self, request: CompletionRequest):
-        _ = self._create(request)
-        for i in _:
-            print(i)
-        # return _
+    # @rcache(noself=True, ttl=15 * 60)
+    @sync_to_async
+    def query(self, q: str):
+        data = list(self.create(q))
+        return {"data": data}
 
-    async def search(self, q: str):
-        return list(self._create(q))
+    def create(self, request: Union[CompletionRequest, str]):
 
-    def _create(self, request: Union[CompletionRequest, str]):
         chunks = zhipuai_sdk_client.assistant.conversation(
 
             assistant_id="659e54b1b8006379b4b2abd6",  # 搜索智能体
@@ -73,14 +71,17 @@ class Completions(object):
         )
 
         references = []
-        buffer = []
         for chunk in chunks:
+            # logger.debug(chunk)
+
             delta = chunk.choices[0].delta
             if hasattr(delta, "tool_calls") and delta.tool_calls:
                 tool_call = delta.tool_calls[0].model_dump()
                 # logger.debug(tool_call)
                 tool_type = tool_call.get("type", "")  # web_browser
                 references += tool_call.get(tool_type, {}).get("outputs") or []  # title link content
+
+                # logger.debug(f"references: {references}")
                 continue
 
             if isinstance(request, CompletionRequest):
@@ -89,16 +90,6 @@ class Completions(object):
                     yield from urls
                     references = []
 
-                # logger.debug(delta)
-                # if delta.content.startswith('【') or buffer: # hasattr(delta, "content")
-                #     buffer.append(delta.content)
-                #     if len(buffer) < 20:
-                #         continue
-                #
-                # if delta.content.endswith('】'):
-                #     delta.content = convert_citations(''.join(buffer))
-                #     if len(buffer) > 25: buffer = []
-
                 delta = chat_completion_chunk.choices[0].delta.model_construct(**delta.model_dump())
                 chat_completion_chunk.choices[0].delta = delta
                 yield chat_completion_chunk
@@ -109,24 +100,21 @@ class Completions(object):
 
 
 if __name__ == '__main__':
-    model = "web-search-pro"
-    # model = "tencent-search"
-
     request = CompletionRequest(
         # model="baichuan4-turbo",
         # model="xx",
         # model="deepseek-r1",
         # model="deepseek-r1:1.5b",
-        model=model,
+        model="model",
 
         # model="moonshot-v1-8k",
         # model="doubao",
 
         messages=[
-            {"role": "user", "content": "《哪吒之魔童闹海》现在的票房是多少"}
+            {"role": "user", "content": "周杰伦"}
         ],
 
         stream=True
     )
     # arun(Completions().search('周杰伦'))
-    arun(Completions().create(request))
+    arun(Completions().query(request.last_user_content))

meutils/caches/common.py

@@ -18,6 +18,8 @@ cache = memory_cache = cached
 
 def rcache(**kwargs):
     """serializer="pickle"
+
+    noself: bool = False,
     :param endpoint: str with the endpoint to connect to. Default is "127.0.0.1".
     :param port: int with the port to connect to. Default is 6379.
     :param db: int indicating database to use. Default is 0.

meutils/data/VERSION

@@ -1 +1 @@
-2025.03.26.18.50.26
+2025.04.01.18.47.48

meutils/io/openai_files.py

@@ -150,9 +150,12 @@ if __name__ == '__main__':
 
     with timer():
         file = "https://top.baidu.com/board?tab=realtime"
-        # file = "https://oss.ffire.cc/files/百炼系列手机产品介绍.docx"
+        file = "http://admin.ilovechatgpt.top/file/yuzhicaizaibutongnianlingquntixiaofeixingweijishichangdiaochawenjuanweishanjianbanpptx_59787479.pptx"
+        file = "https://oss.ffire.cc/files/百炼系列手机产品介绍.docx"
         # file = "https://app.yinxiang.com/fx/8b8bba1e-b254-40ff-81e1-fa3427429efe"
 
+        print(guess_mime_type(file))
+
         arun(file_extract(file))
 
         # arun(file_extract("/Users/betterme/PycharmProjects/AI/data/041【精选】海门招商重工5G+智慧工厂解决方案.pptx"))

meutils/llm/check_utils.py

@@ -146,7 +146,7 @@ if __name__ == '__main__':
 
     # arun(check_token_for_jina(["jina_c8da77fed9704d558c8def39837960edplTLkNYrsPTJHBF1HcYg_RkRVh0X"]*10))
 
-    # arun(check_token_for_siliconflow("sk-zdsfykbjrdzobopoienanvlpqdsttbbmgpayctqokkrwqazg"))
+    arun(check_token_for_siliconflow("sk-vpeietyomqjvizlzfuztzthggcqvutowgbmhjggsmwuhsomg"))
     "https://xchatllm.feishu.cn/sheets/Bmjtst2f6hfMqFttbhLcdfRJnNf?sheet=79272d"
 
-    arun(check_token_for_moonshot("sk-Qnr87vtf2Q6MEfc2mVNkVZ4qaoZg3smH9527I25QgcFe7HrT"))
+    # arun(check_token_for_moonshot("sk-Qnr87vtf2Q6MEfc2mVNkVZ4qaoZg3smH9527I25QgcFe7HrT"))

meutils/llm/clients.py

@@ -33,19 +33,17 @@ zhipuai_sdk_client = ZhipuAI(
     base_url=os.getenv("ZHIPUAI_BASE_URL")
 )
 
-
-
-
-qwen_client = AsyncOpenAI(
-    base_url="https://all.chatfire.cn/qwen/v1",
-    api_key="eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpZCI6IjMxMGNiZGFmLTM3NTQtNDYxYy1hM2ZmLTllYzgwMDUzMjljOSIsImV4cCI6MTc0MzAzNTk4OH0.GVAoSFtK94a9CgxqHCEnxzAnRi7gafIvYyH9mIJUh4s"
-)
-
 # zhipuai_client = OpenAI(
 #     api_key=os.getenv("ZHIPUAI_API_KEY"),
 #     base_url=os.getenv("ZHIPUAI_BASE_URL")
 # )
 
+# ark_bots_client = AsyncOpenAI(
+#     api_key=os.getenv("ZHIPUAI_API_KEY"),
+#     base_url="https://ark.cn-beijing.volces.com/api/v3/bots"
+# )
+
+
 if __name__ == '__main__':
     from meutils.pipe import *
 

meutils/llm/completions/chat_gemini.py

@@ -33,7 +33,7 @@ class Completions(object):
                  api_key: Optional[str] = None
                  ):
 
-        base_url = "https://api.aiguoguo199.com/v1"
+        base_url = "https://api0.aiguoguo199.com/v1"
 
         self.client = AsyncOpenAI(
             base_url=base_url,

meutils/llm/completions/deep2x.py

@@ -20,7 +20,6 @@ from meutils.io.files_utils import to_bytes
 from meutils.io.openai_files import file_extract, guess_mime_type
 from meutils.str_utils.json_utils import json_path
 from meutils.apis.search import metaso
-# from meutils.apis.chatglm import glm_video_api
 
 from meutils.llm.clients import chatfire_client, zhipuai_client, AsyncOpenAI
 from meutils.llm.openai_utils import to_openai_params
@@ -28,7 +27,12 @@ from meutils.llm.openai_utils import to_openai_params
 from meutils.schemas.openai_types import ChatCompletionRequest
 from meutils.schemas.openai_types import chat_completion, chat_completion_chunk, CompletionRequest, ImageRequest
 
-
+"""
+delta = chunk.choices[0].delta
+| │ └ []
+| └ ChatCompletionChunk(id='02174299556532927e9140493fc1cd076b4fe1b883ff101a83257', choices=[], created=1742995565, model='deep-d...
+|
+"""
 class Completions(object):
 
     def __init__(self, api_key: Optional[str] = None):
@@ -47,10 +51,14 @@ class Completions(object):
             reasoning_content = ""
             completions = await chatfire_client.chat.completions.create(**data)
             async for chunk in completions:
-                yield chunk
-                delta = chunk.choices[0].delta
-                if hasattr(delta, "reasoning_content"):
-                    reasoning_content += delta.reasoning_content
+                if chunk.choices:  # 自定义没问题，todo:openai通道报错
+                    # logger.debug(chunk)
+                    yield chunk
+                    delta = chunk.choices[0].delta
+                    if hasattr(delta, "reasoning_content"):
+                        reasoning_content += delta.reasoning_content
+                else:
+                    logger.error(chunk)
 
             request.messages = [
                 {
@@ -58,8 +66,10 @@ class Completions(object):
                     'content': f"""<think>\n\n{reasoning_content}\n\n</think>\n\n{request.last_user_content}"""
                 }
             ]
+            logger.debug(request)
             data = to_openai_params(request)
             async for chunk in await self.client.chat.completions.create(**data):
+                # logger.debug(chunk)
                 yield chunk
         else:
             reasoning_content = ""

meutils/llm/completions/google2openai.py

@@ -0,0 +1,75 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+# @Project      : AI.  @by PyCharm
+# @File         : google2openai
+# @Time         : 2025/4/1 13:38
+# @Author       : betterme
+# @WeChat       : meutils
+# @Software     : PyCharm
+# @Description  : 
+"""
+1. 生图 编辑图 多轮
+2. 文件上传、大文件问答
+
+注意 ：借助 File API，您最多可为每个项目存储 20 GB 的文件，每个文件的大小上限为 2 GB。文件会存储 48 小时。您可以在该时间段内使用 API 密钥访问这些数据，但无法从 API 下载这些数据。在已推出 Gemini API 的所有地区，此功能均可免费使用。
+
+"""
+from meutils.pipe import *
+
+from google import genai
+from google.genai.types import HttpOptions, GenerateContentConfig, Content, HarmCategory, HarmBlockThreshold, Part
+
+# Content(role="user", parts=[Part.from_text(text=prompt)]),
+# Content(role="model", parts=[Part.from_text(text="Ok")]),
+
+config = GenerateContentConfig(
+
+    temperature=0.7,
+    top_p=0.8,
+    # response_modalities=['Text', 'Image'],
+
+    # 公民诚信类别的默认屏蔽阈值为 Block none（对于别名为 gemini-2.0-flash、gemini-2.0-pro-exp-02-05 和 gemini-2.0-flash-lite 的 gemini-2.0-flash-001），适用于 Google AI Studio 和 Gemini API；仅适用于 Google AI Studio 中的所有其他模型的 Block most。
+    # safety_settings=[
+    #     SafetySetting(
+    #         category=HarmCategory.HARM_CATEGORY_HATE_SPEECH,
+    #         threshold=HarmBlockThreshold.BLOCK_LOW_AND_ABOVE,
+    #     ),
+    # ]
+)
+
+# self._http_options.base_url = 'https://generativelanguage.googleapis.com/'
+# self._http_options.api_version = 'v1beta'
+client = genai.Client(
+    api_key="AIzaSyD19pv1qsYjx4ZKbfH6qvNdYzHMV2TxmPU",
+    http_options=HttpOptions(
+        base_url="https://all.chatfire.cc/genai"
+    )
+)
+
+file = "/Users/betterme/PycharmProjects/AI/QR.png"
+#
+# file_object = client.files.upload(file=file)
+# prompt = "一句话总结"
+
+file_object = client.aio.files.upload(file=file)
+# https://generativelanguage.googleapis.com/v1beta/files/ickgffcfb9zl
+#
+# contents = ('Hi, can you create a 3d rendered image of a pig '
+#             'with wings and a top hat flying over a happy '
+#             'futuristic scifi city with lots of greenery?')
+#
+# prompt = "9.11 9.8哪个大呢"
+#
+response = client.models.generate_content(
+    # model="gemini-2.0-flash-exp-image-generation",
+
+    model = "gemini-2.5-pro-exp-03-25",
+
+    contents=prompt,
+    # config=config
+)
+# client.aio.
+# client.aio.chats.create()
+
+if __name__ == '__main__':
+    arun(file_object)

meutils/llm/completions/qwenllm.py

@@ -21,7 +21,6 @@ from meutils.decorators.retry import retrying
 from meutils.io.files_utils import to_bytes, guess_mime_type
 from meutils.caches import rcache
 
-from meutils.llm.clients import qwen_client
 from meutils.llm.openai_utils import to_openai_params
 
 from meutils.config_utils.lark_utils import get_next_token_for_polling
@@ -38,8 +37,15 @@ cookie = "_gcl_au=1.1.1784604298.1740443944;xlly_s=1;isg=BNHREY3fUiqB6r5JMj56XbB
 
 
 @retrying()
-@rcache(ttl=3600, serializer='pickle')
-async def to_file(file):
+async def to_file(file, api_key):
+    qwen_client = AsyncOpenAI(
+        base_url="https://all.chatfire.cn/qwen/v1",
+        api_key=api_key,
+        default_headers={
+            'User-Agent': ua.random,
+            'Cookie': cookie
+        }
+    )
     filename = Path(file).name if isinstance(file, str) else 'untitled'
     mime_type = guess_mime_type(file)
     file_bytes: bytes = await to_bytes(file)
@@ -71,7 +77,7 @@ async def create(request: CompletionRequest, token: Optional[str] = None):  # Ch
         request.model = "qwen-max-latest"
         request.messages[-1]['chat_type'] = "search"
 
-    if any(i in model for i in ("qwq", "think")):  # qwq-max-search
+    if any(i in model for i in ("qwq", "qvq", "think", "thinking")):  # qwq-max-search
         request.model = "qwen-max-latest"
         request.messages[-1]['feature_config'] = {"thinking_enabled": True}
 
@@ -87,13 +93,13 @@ async def create(request: CompletionRequest, token: Optional[str] = None):  # Ch
             for i, content in enumerate(user_content):
                 if content.get("type") == 'file_url':  # image_url file_url video_url
                     url = content.get(content.get("type")).get("url")
-                    file_object = await to_file(url)
+                    file_object = await to_file(url, client)
 
                     user_content[i] = {"type": "file", "file": file_object.id}
 
                 elif content.get("type") == 'image_url':
                     url = content.get(content.get("type")).get("url")
-                    file_object = await to_file(url)
+                    file_object = await to_file(url, client)
 
                     user_content[i] = {"type": "image", "image": file_object.id}
 
@@ -102,7 +108,7 @@ async def create(request: CompletionRequest, token: Optional[str] = None):  # Ch
 
             user_content = [{"type": "text", "text": user_content}]
 
-            file_object = await to_file(file_url)
+            file_object = await to_file(file_url, client)
 
             content_type = file_object.meta.get("content_type", "")
             if content_type.startswith("image"):
@@ -130,13 +136,6 @@ async def create(request: CompletionRequest, token: Optional[str] = None):  # Ch
         if not isinstance(response, str):
             yield response.choices[0].message.content  # isinstance(response, str)
 
-        # for i in range(3):
-        #     if not isinstance(response, str):  # 报错
-        #         yield response.choices[0].message.content
-        #         break
-        #     else:
-        #         logger.warning(f"重试 {i}\n{response}")
-
 
 if __name__ == '__main__':
     # [
@@ -150,36 +149,39 @@ if __name__ == '__main__':
     #     "qwen2.5-32b-instruct"
     # ]
 
-    # user_content = [
-    #     {
-    #         "type": "text",
-    #         "text": "解读图片"
-    #     },
-    #     {
-    #         "type": "image_url",
-    #         "image_url": {
-    #             "url": "https://fyb-pc-static.cdn.bcebos.com/static/asset/homepage@2x_daaf4f0f6cf971ed6d9329b30afdf438.png"
-    #         }
-    #     }
-    # ]
-
     user_content = [
         {
             "type": "text",
-            "text": "总结下"
+            "text": "一句话总结"
         },
         {
-            "type": "file_url",
-            "file_url": {
-                "url": "https://oss.ffire.cc/files/AIGC.pdf"
+            "type": "image_url",
+            "image_url": {
+                "url": "https://fyb-pc-static.cdn.bcebos.com/static/asset/homepage@2x_daaf4f0f6cf971ed6d9329b30afdf438.png"
             }
         }
-
     ]
 
+    # user_content = [
+    #     {
+    #         "type": "text",
+    #         "text": "总结下"
+    #     },
+    #     {
+    #         "type": "file_url",
+    #         "file_url": {
+    #             "url": "https://oss.ffire.cc/files/AIGC.pdf"
+    #         }
+    #     }
+    #
+    # ]
+
     request = CompletionRequest(
         # model="qwen-turbo-2024-11-01",
-        model="qwen-max-latest",
+        # model="qwen-max-latest",
+        # model="qvq-max-2025-03-25",
+        model="qvq-72b-preview-0310",
+
         # model="qwen-max-latest-search",
         # model="qwq-max",
         # model="qwq-32b-preview",
@@ -195,12 +197,12 @@ if __name__ == '__main__':
             {
                 'role': 'user',
                 # 'content': '今天南京天气',
-                'content': "9.8 9.11哪个大",
+                # 'content': "9.8 9.11哪个大",
                 # 'content': 'https://oss.ffire.cc/files/AIGC.pdf 总结下',
 
                 # "chat_type": "search",
 
-                # 'content': user_content,
+                'content': user_content,
 
                 # "content": [
                 #     {

meutils/llm/openai_utils/common.py

@@ -62,6 +62,11 @@ def to_openai_params(
     data['extra_body'] = extra_body  # 拓展字段
     data['model'] = redirect_model or data['model']
 
+    if request.model.startswith(("gemini",)):
+        data.pop("extra_body", None)
+        data.pop("presence_penalty", None)
+        data.pop("frequency_penalty", None)
+
     return data
 
 

meutils/schemas/oneapi/common.py

@@ -248,6 +248,7 @@ MODEL_PRICE = {
 
     "gpt-4-all": 0.1,
     "gpt-4o-all": 0.1,
+    "gpt-4o-image": 0.06,
 
     "gpt-4-gizmo-*": 0.1,
     "advanced-voice": 1,
@@ -427,6 +428,10 @@ MODEL_RATIO = {
     "qwen-vl-max-latest": 1.5,
     "qwen-vl-plus-latest": 0.75,
 
+    "qwen2.5-vl-7b-instruct": 0.15,
+    "qwen2.5-vl-32b-instruct": 0.5,
+    "qwen2.5-vl-72b-instruct": 1.5,
+
     "qwen2.5-coder-7b-instruct": 0.05,
     "qwen2.5-7b-instruct": 0.05,
     "qwen2.5-14b-instruct": 0.25,
@@ -436,12 +441,13 @@ MODEL_RATIO = {
     "qwen2.5-coder-32b-instruct": 0.5,
 
     "qwq-32b": 1,
-    "qwq-max": 2,
+    "qwq-plus": 0.8,
+    "qwq-max": 0.8,
     "qwq-max-search": 2,
     "qwen-max-search": 2,
 
     "qvq-72b-preview": 2,
-    "qwen2.5-vl-72b-instruct": 8,
+    "qvq-max-2025-03-25": 4,
 
     "qwen1.5-7b-chat": 0.05,  # 特价
     "qwen1.5-14b-chat": 0.7,
@@ -904,6 +910,11 @@ COMPLETION_RATIO = {
     "qwen-max": 4,
     "qwen-vl-max-latest": 3,
     "qwen-vl-plus-latest": 3,
+
+    "qwen2.5-vl-7b-instruct": 4,
+    "qwen2.5-vl-32b-instruct": 4,
+    "qwen2.5-vl-72b-instruct": 4,
+
     "qwen2-vl-7b-instruct": 5,
     "qwen2-vl-72b-instruct": 5,
     "qwen-max-latest": 4,
@@ -912,16 +923,17 @@ COMPLETION_RATIO = {
 
     "qwen-plus": 2.5,
 
-    "qwq-max": 4,
+    "qwq-plus": 2.5,
+    "qwq-max": 2.5,
     "qwq-max-search": 4,
     "qwen-max-search": 4,
+    "qvq-max-2025-03-25": 4,
 
     "qwen2.5-7b-instruct": 4,
     "qwen2.5-14b-instruct": 4,
     "qwen2.5-32b-instruct": 4,
     "qwen2.5-72b-instruct": 4,
     "qwen2.5-math-72b-instruct": 4,
-    "qwen2.5-vl-72b-instruct": 3,
 
     "deepseek-vl2": 4,
     "deepseek-ai/deepseek-vl2": 4,