MeUtils 2025.1.29.10.0.47__py3-none-any.whl → 2025.2.6.13.5.49__py3-none-any.whl

meutils/llm/clients.py

@@ -14,6 +14,9 @@ from meutils.pipe import *
 OpenAI = lru_cache(Client)
 AsyncOpenAI = lru_cache(AsyncClient)
 
+
+chatfire_client = AsyncOpenAI()
+
 moonshot_client = AsyncOpenAI(
     api_key=os.getenv("MOONSHOT_API_KEY"),
     base_url=os.getenv("MOONSHOT_BASE_URL")

meutils/llm/completions/agents/search.py

@@ -9,7 +9,7 @@
 # @Description  : todo: 格式匹配
 
 from meutils.pipe import *
-from meutils.llm.clients import AsyncOpenAI, zhipuai_client, moonshot_client
+from meutils.llm.clients import AsyncOpenAI, chatfire_client, zhipuai_client, moonshot_client
 from meutils.llm.openai_utils import to_openai_params
 from meutils.schemas.openai_types import chat_completion, chat_completion_chunk, ChatCompletionRequest, CompletionUsage
 
@@ -90,18 +90,33 @@ class Completions(object):
             completion = await client.chat.completions.create(**data)
             return completion
 
-        else:  # glm
-            request.model = "web-search-pro"
+        else:
+            # 搜索
             data = to_openai_params(request)
-            completion = await zhipuai_client.chat.completions.create(**data)
-            logger.debug(completion)
+            data['model'] = "web-search-pro"
+            data['stream'] = False
+            search_completion = await zhipuai_client.chat.completions.create(**data)
+            logger.debug(search_completion)
+
+            # 大模型
+            request.messages.append({
+                "role": "user",
+                "content": search_completion.model_dump_json(indent=4),
+            })
+
+            data = to_openai_params(request)
+
+            completion = await chatfire_client.chat.completions.create(**data)
             return completion
 
 
 if __name__ == '__main__':
     request = ChatCompletionRequest(
         # model="baichuan4-turbo",
-        model="xx",
+        # model="xx",
+        # model="deepseek-r1",
+        model="deepseek-r1:1.5b",
+
         # model="moonshot-v1-8k",
         # model="doubao",
 
@@ -113,3 +128,10 @@ if __name__ == '__main__':
     )
 
     arun(Completions().create(request))
+
+    # async def test():
+    #     for i in await Completions().create(request):
+    #         print(i)
+    #
+    #
+    # arun(test())

meutils/llm/completions/qwenllm.py

@@ -8,8 +8,9 @@
 # @Software     : PyCharm
 # @Description  :
 
+from openai import AsyncOpenAI
+
 from meutils.pipe import *
-from meutils.llm.clients import AsyncOpenAI
 from meutils.llm.openai_utils import to_openai_params
 
 from meutils.config_utils.lark_utils import get_next_token_for_polling
@@ -19,11 +20,18 @@ FEISHU_URL = "https://xchatllm.feishu.cn/sheets/Bmjtst2f6hfMqFttbhLcdfRJnNf?shee
 
 base_url = "https://chat.qwenlm.ai/api"
 
+from fake_useragent import UserAgent
+
+ua = UserAgent()
+
 
 async def create(request: ChatCompletionRequest):
     token = await get_next_token_for_polling(feishu_url=FEISHU_URL)
 
-    client = AsyncOpenAI(base_url=base_url, api_key=token)
+    client = AsyncOpenAI(
+        base_url=base_url, api_key=token,
+        default_headers={'User-Agent': ua.random}
+    )
     data = to_openai_params(request)
 
     if request.stream:
@@ -35,6 +43,7 @@ async def create(request: ChatCompletionRequest):
 
     else:
         response = await client.chat.completions.create(**data)
+        # logger.info(response)
         yield response.choices[0].message.content
 
 
@@ -51,7 +60,7 @@ if __name__ == '__main__':
     # ]
     request = ChatCompletionRequest(
         # model="qwen-turbo-2024-11-01",
-        model="qwen-turbo-latest",
+        model="qwen-max-latest",
         # model="qwen-plus-latest",
 
         messages=[

meutils/llm/completions/reasoner.py

@@ -0,0 +1,87 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+# @Project      : AI.  @by PyCharm
+# @File         : reasoner
+# @Time         : 2025/2/6 08:35
+# @Author       : betterme
+# @WeChat       : meutils
+# @Software     : PyCharm
+# @Description  : 
+
+
+from openai import AsyncOpenAI
+
+from meutils.pipe import *
+from meutils.llm.clients import chatfire_client, AsyncOpenAI
+from meutils.llm.openai_utils import to_openai_params
+
+from meutils.schemas.openai_types import chat_completion, chat_completion_chunk, ChatCompletionRequest, CompletionUsage
+
+
+class Completions(object):
+
+    def __init__(self, api_key: Optional[str] = None, base_url: Optional[str] = None):
+        self.api_key = api_key
+        self.base_url = base_url
+
+        self.client = AsyncOpenAI(
+            base_url=self.base_url, api_key=self.api_key,
+        )
+
+    async def create(self, request: ChatCompletionRequest):
+        data = to_openai_params(request)
+
+        if request.stream:
+            _chunk = ""
+            async for chunk in await self.client.chat.completions.create(**data):
+                delta = chunk.choices[0].delta
+
+                reasoning_content = "> Reasoning\n"  # "> Reasoning\n"  # 前缀 "> 思考中\n"
+                if hasattr(delta, 'reasoning_content'):
+                    if reasoning_content:
+                        reasoning_content += delta.reasoning_content
+                    yield reasoning_content
+                    reasoning_content = ''
+                else:
+                    yield '\n'
+
+                yield delta.content
+
+        else:
+            completions = await self.client.chat.completions.create(**data)
+            yield completions
+
+
+if __name__ == '__main__':
+    # [
+    #     "qwen-plus-latest",
+    #     "qvq-72b-preview",
+    #     "qwq-32b-preview",
+    #     "qwen2.5-coder-32b-instruct",
+    #     "qwen-vl-max-latest",
+    #     "qwen-turbo-latest",
+    #     "qwen2.5-72b-instruct",
+    #     "qwen2.5-32b-instruct"
+    # ]
+    request = ChatCompletionRequest(
+        # model="qwen-turbo-2024-11-01",
+        # model="qwen-max-latest",
+        # model="qwen-plus-latest",
+
+        model="deepseek-r1:1.5b",
+        # model="deepseek-r1",
+
+        messages=[
+            {
+                'role': 'system',
+                'content': '每次回答前务必思考之后m，回答问题'
+            },
+            {
+                'role': 'user',
+                'content': '你好'
+            },
+
+        ],
+        stream=True,
+    )
+    arun(Completions().create(request))

meutils/llm/openai_utils/common.py

@@ -282,3 +282,5 @@ if __name__ == '__main__':
     arun(main())
 
     # to_openai_params(ChatCompletionRequest())
+
+    print(token_encoder.encode('hi'))

meutils/schemas/metaso_types.py

@@ -15,11 +15,13 @@ FEISHU_URL = "https://xchatllm.feishu.cn/sheets/Bmjtst2f6hfMqFttbhLcdfRJnNf?shee
 
 
 class MetasoRequest(BaseModel):
-    question: str = "Chatfire"
+    model: Optional[Literal["ds-r1",]] = None
 
     """search-mini search search-pro"""
     mode: Literal["concise", "detail", "research"] = "detail"  # concise detail research
 
+    question: str = "Chatfire"
+
     """全网 文库 学术 图片 播客"""
     engineType: str = ""  # scholar
 
@@ -28,6 +30,10 @@ class MetasoRequest(BaseModel):
     searchTopicId: Optional[str] = None
     searchTopicName: Optional[str] = None
 
+    enableMix: bool = True
+    newEngine: bool = True
+    enableImage: bool = True
+
     # 自定义字段
     response_format: Optional[str] = None  # 原生内容
 

meutils/schemas/oneapi/common.py

@@ -61,7 +61,7 @@ MODEL_PRICE = {
 
     "official-api-vidu": 0.5,
 
-    "official-api-hailuo-video": 0.3,
+    "official-api-hailuo-video": 0.5,
     "api-hailuo-video": 0.3,
 
     # delle3
@@ -107,6 +107,8 @@ MODEL_PRICE = {
 
     "api-hunyuan-video": 0.1,
 
+    "deepseek-ai/Janus-Pro-7B": 0.01,
+
     # replicate
     "api-replicate-flux-1.1-pro": 0.040,
     "api-replicate-flux-1.1-pro-ultra": 0.060,
@@ -215,6 +217,9 @@ MODEL_PRICE = {
     "o1-mini-all": 0.2,
     "o1-preview-all": 0.6,
 
+    "o3-mini": 0.05,
+    "o3-mini-high": 0.1,
+
     "gpt-4-all": 0.1,
     "gpt-4o-all": 0.1,
 
@@ -377,16 +382,17 @@ MODEL_RATIO = {
     "qwen-max": 10,
     "qwen-max-longcontext": 20,
     "qwen-turbo-2024-11-01": 0.15,
+    "qwen-max-latest":0.01,
 
     "qwen-vl-max-latest": 1.5,
     "qwen-vl-plus-latest": 0.75,
 
     "qwen2.5-coder-7b-instruct": 0.05,
     "qwen2.5-7b-instruct": 0.05,
-    "qwen2.5-14b-instruct": 0.7 / 2,
-    "qwen2.5-32b-instruct": 1.26 / 2,
-    "qwen2.5-72b-instruct": 4.13 / 2,
-    "qwen2.5-math-72b-instruct": 4.13 / 2,
+    "qwen2.5-14b-instruct": 0.25,
+    "qwen2.5-32b-instruct": 0.5,
+    "qwen2.5-72b-instruct": 2,
+    "qwen2.5-math-72b-instruct": 2,
     "qwen2.5-coder-32b-instruct": 0.5,
     "qwen/qwq-32b-preview": 1,
     "Qwen/QwQ-32B-Preview": 1,
@@ -455,7 +461,12 @@ MODEL_RATIO = {
 
     # deepseek
     "deepseek-chat": 0.5,
+    "deepseek-vl2": 0.5,
+
     "deepseek-ai/deepseek-vl2": 0.5,
+    "deepseek-llm-67b-chat": 0.5,
+
+    "deepseek-chat:function": 4,
 
     "deepseek-v3": 0.5,
     "deepseek/deepseek-chat": 0.5,
@@ -464,10 +475,14 @@ MODEL_RATIO = {
 
     "deepseek-r1:1.5b": 0.1,
     "deepseek-r1:7b": 0.2,
+    "deepseek-r1:14b": 0.5,
+    "deepseek-r1:32b": 1,
     'deepseek-r1': 2,
     'deepseek-reasoner': 2,
     'deepseek-think': 0.5,
     "deepseek-search": 0.5,
+    "deepseek-chat-64k": 5,
+    "deepseek-v3-128k": 5,
 
     # 豆包
     "doubao-lite-128k": 0.4,
@@ -618,6 +633,7 @@ MODEL_RATIO = {
     "o1-preview": 7.5,
     "o1-mini-2024-09-12": 1.5,
     "o1-preview-2024-09-12": 7.5,
+    "o3-mini": 0.55,
 
     # 硅基
     "llama-3.1-8b-instruct": 0.01,
@@ -684,10 +700,13 @@ COMPLETION_RATIO = {
     "gpt-4-all": 4,
     "gpt-4-gizmo-*": 4,
     "gpt-4o-all": 4,
+
     "o1-mini": 4,
     "o1-preview": 4,
     "o1-mini-2024-09-12": 4,
     "o1-preview-2024-09-12": 4,
+    "o3-mini": 4,
+
     "gpt-4o-realtime-preview": 4,
     "gpt-4o-realtime-preview-2024-10-01": 4,
     "gpt-4o-2024-11-20": 4,
@@ -753,8 +772,16 @@ COMPLETION_RATIO = {
     "qwen-vl-plus-latest": 3,
     "qwen2-vl-7b-instruct": 5,
     "qwen2-vl-72b-instruct": 5,
+    "qwen-max-latest": 5,
+
+    "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,
 
-    "deepseek-ai/deepseek-vl2": 3,
+    "deepseek-vl2": 4,
+    "deepseek-ai/deepseek-vl2": 4,
 
     # 豆包
     "doubao-lite-128k": 3,
@@ -775,13 +802,17 @@ COMPLETION_RATIO = {
     "doubao-vision-lite-32k": 3,
     "doubao-vision-pro-32k": 3,
 
-    "deepseek-v3": 1,
+    "deepseek-v3": 4,
     "deepseek-search": 1,
     'deepseek-r1': 4,
     'deepseek-reasoner': 4,
     "deepseek/deepseek-chat": 1,
     "deepseek-ai/DeepSeek-V3": 1,
     "accounts/fireworks/models/deepseek-v3": 1,
+    "deepseek-chat:function": 4,
+    "deepseek-chat-64k": 5,
+    "deepseek-v3-128k": 5,
+    "deepseek-llm-67b-chat": 4,
 
     "glm-zero": 5,
     "glm-zero-preview": 5,

meutils/schemas/vidu_types.py

@@ -147,7 +147,7 @@ class ViduRequest(BaseModel):
                 "input": input,
                 "type": self.type or type,
                 "settings": {
-                    "model_version": "2.0",
+                    "model_version": "2.0", #######
                     "style": "general",
                     "duration": self.duration,