MeUtils 2025.3.14.8.43.3__py3-none-any.whl → 2025.3.19.19.13.35__py3-none-any.whl

examples/_openaisdk/zhipu_/346/231/272/350/203/275/344/275/223.py

@@ -9,21 +9,28 @@
 # @Description  : https://bigmodel.cn/dev/api/intelligent-agent-model/assistantapi
 
 from meutils.pipe import *
+from meutils.llm.clients import zhipuai_sdk_client
 
 from zhipuai import ZhipuAI
-api_key = "YOUR API KEY"
-url = "https://open.bigmodel.cn/api/paas/v4"
-client = ZhipuAI(
-    api_key="e21bd630f681c4d90b390cd609720483.WSFVgA3KkwNCX0mN",
-    base_url="https://open.bigmodel.cn/api/paas/v4/"
-)
 
-print(bjson(client.assistant.query_support()))
-generate = client.assistant.conversation(
+# api_key = "YOUR API KEY"
+# url = "https://open.bigmodel.cn/api/paas/v4"
+# client = ZhipuAI(
+#     api_key="e21bd630f681c4d90b390cd609720483.WSFVgA3KkwNCX0mN",
+#     base_url="https://open.bigmodel.cn/api/paas/v4/"
+# )
+
+
+#
+#
+# # print(bjson(client.assistant.query_support()))
+generate = zhipuai_sdk_client.assistant.conversation(
     # assistant_id="65a265419d72d299a9230616",
-    # assistant_id="65940acff94777010aa6b796",
-    assistant_id="676411c38945bbc58a905d31",
-    conversation_id=None,
+    # assistant_id="659d051a5f14eb8ce1235b96",
+    assistant_id="65d2f07bb2c10188f885bd89",
+
+    # assistant_id="659e54b1b8006379b4b2abd6",
+    conversation_id=None,  # 多轮
     model="glm-4-assistant",
     messages=[
         {
@@ -31,7 +38,8 @@ generate = client.assistant.conversation(
             "content": [{
                 "type": "text",
                 # "text": "北京未来七天气温，做个折线图",
-                "text": "画条狗"
+                # "text": "画条狗"
+                "text": "周杰伦",
 
             }]
         }
@@ -42,4 +50,59 @@ generate = client.assistant.conversation(
 )
 
 for resp in generate:
-    print(resp)
+    delta = resp.choices[0].delta
+    if hasattr(delta, "tool_calls") and delta.tool_calls:
+        # print(delta)
+
+        print(delta.model_dump_json(indent=4, exclude_none=True))
+
+    # print(delta.model_dump_json(indent=4, exclude_none=True))
+    # for tc in delta.tool_calls:
+    #     tc.web_browser
+
+    # print(resp.choices[0])
+
+    # if hasattr(delta, "content"):
+    #     print(delta.content)
+
+    # 输入 输出
+    # input
+# {
+#     "assistant_id": "65d2f07bb2c10188f885bd89",
+#     "conversation_id": "67d932d5e579c3ded42aa80e",
+#     "meta_data": {
+#         "if_plus_model": false,
+#         "is_test": false,
+#         "input_question_type": "xxxx",
+#         "channel": "",
+#         "draft_id": "",
+#         "quote_log_id": "",
+#         "platform": "pc"
+#     },
+#     "messages": [
+#         {
+#             "role": "user",
+#             "content": [
+#                 {
+#                     "type": "text",
+#                     "text": "基于这个内容做个ppt"
+#                 },
+#                 {
+#                     "type": "file",
+#                     "file": [
+#                         {
+#                             "file_id": "chatglm4/3db10f79-a952-4987-83d2-cf0cfd5d5530.pdf",
+#                             "file_url": "https://sfile.chatglm.cn/chatglm4/3db10f79-a952-4987-83d2-cf0cfd5d5530.pdf",
+#                             "file_name": "附件.大模型在合规管理工作中的应用.pdf",
+#                             "file_size": 2571523,
+#                             "order": 0,
+#                             "maxReadPercent": 0,
+#                             "cover_images": [],
+#                             "url": "https://sfile.chatglm.cn/chatglm4/3db10f79-a952-4987-83d2-cf0cfd5d5530.pdf"
+#                         }
+#                     ]
+#                 }
+#             ]
+#         }
+#     ]
+# }

meutils/apis/jimeng/common.py

@@ -180,6 +180,8 @@ if __name__ == '__main__':
     token = "34438eb03d165737122180caf62a8058"
     token = "ffeee346fbd19eceebb79a7bfbca4bfe"
     token = "b8bb4cb67dba6c0d1048bdc0596bc461"
+    token = "34438eb03d165737122180caf62a8058"
+    token = "a521dd578adcfb191fad38dd4baab498"
     arun(check_token(token))
 
     # print(arun(aget_spreadsheet_values(feishu_url=FEISHU_URL, to_dataframe=True))[0].tolist())

meutils/apis/jimeng/images.py

@@ -272,8 +272,8 @@ async def get_task(task_id, token):
         logger.debug(f"status: {status_code}")
 
         # 敏感词存储
-        if status_code != 50:
-            send_message_for_images(task_info, __name__)
+        # if status_code != 50:
+        #     send_message_for_images(task_info, __name__)
 
         """
         "status": 30, # 内容审核
@@ -376,10 +376,10 @@ if __name__ == '__main__':
     # arun(generate(ImageRequest(**data)))
 
     # arun(generate(ImageRequest(prompt="fuck you")))
-    # prompt = "A plump Chinese beauty wearing a wedding dress revealing her skirt and underwear is swinging on the swing,Happy smile,cleavage,Exposed thighs,Spread your legs open,Extend your leg,panties,upskirt,Barefoot,sole"
-    # request = ImageRequest(prompt=prompt)
-    task = arun(create_task(ImageRequest(**data), token=token))
+    prompt = "A plump Chinese beauty wearing a wedding dress revealing her skirt and underwear is swinging on the swing,Happy smile,cleavage,Exposed thighs,Spread your legs open,Extend your leg,panties,upskirt,Barefoot,sole"
+    request = ImageRequest(prompt=prompt)
+    # task = arun(create_task(ImageRequest(**data), token=token))
 
     # arun(get_task(task.task_id, task.system_fingerprint))
 
-    # arun(generate(request))
+    arun(generate(request))

meutils/apis/jina/__init__.py

@@ -0,0 +1,11 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+# @Project      : AI.  @by PyCharm
+# @File         : __init__.py
+# @Time         : 2025/3/17 18:30
+# @Author       : betterme
+# @WeChat       : meutils
+# @Software     : PyCharm
+# @Description  : 
+
+from meutils.apis.jina.common import url_reader

meutils/apis/jina/common.py

@@ -0,0 +1,43 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+# @Project      : AI.  @by PyCharm
+# @File         : common
+# @Time         : 2025/3/17 18:44
+# @Author       : betterme
+# @WeChat       : meutils
+# @Software     : PyCharm
+# @Description  :
+
+from httpx import AsyncClient
+
+from meutils.pipe import *
+from meutils.decorators.retry import retrying
+from meutils.caches import cache, rcache
+
+
+@retrying()
+@rcache(ttl=300)
+async def _url_reader(url: str):
+    """markdown"""
+
+    async with AsyncClient(base_url="https://r.jina.ai", headers={}) as client:
+        response = await client.get(f"/{url}")
+        response.raise_for_status()
+
+        return response.text
+
+
+async def url_reader(urls: Union[str, List[str]]):
+    if isinstance(urls, str):
+        return await _url_reader(urls)
+
+    tasks = [_url_reader(url) for url in urls]
+    return await asyncio.gather(*tasks)
+
+
+if __name__ == '__main__':
+    url = "https://top.baidu.com/board?tab=realtime"
+
+    # print(arun(url_reader(url)))
+
+    print(arun(url_reader([url])))

meutils/apis/oneapi/channel.py

@@ -16,6 +16,7 @@ async def edit_channel(models, token: Optional[str] = None):
     token = token or os.environ.get("CHATFIRE_ONEAPI_TOKEN")
 
     models = ','.join(filter(lambda model: model.startswith(("api", "official-api", "ppu", "kling-v")), models))
+    models += ",suno-v3"
 
     payload = {
         "id": 289,
@@ -41,8 +42,8 @@ async def edit_channel(models, token: Optional[str] = None):
         "auto_ban": 0,
         "other_info": "",
 
-        "group": "default,chatfire,vip3",  # ','.join(GROUP_RATIO),
-        "groups": ['default', 'chatfire', 'vip3', 'vip8', '2B']
+        "group": "default,china",  # ','.join(GROUP_RATIO),
+        "groups": ['default']
     }
     headers = {
         'authorization': f'Bearer {token}',

meutils/apis/oneapi/user.py

@@ -113,4 +113,4 @@ if __name__ == '__main__':
 
     # arun(get_api_key_log('sk-gpoH1z3G6nHovD8MY40i6xx5tsC1vbh7B3Aao2jmejYNoKhv'))
     # arun(get_user_money("sk-LlB4W38z9kv5Wy1c3ceeu4PHeIWs6bbWsjr8Om31jYvsucRv"))
-    arun(get_user_from_api_key('sk-gpoH1z3G6nHovD8MY40i6xx5tsC1vbh7B3Aao2jmejYNoKhv'))
+    arun(get_user_from_api_key('sk-hXruPMHP0TnBpohQMJWBZILDVKugUmzQF0lwZTetDX7eHaAM'))

meutils/apis/search/_web_search.py

@@ -0,0 +1,87 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+# @Project      : AI.  @by PyCharm
+# @File         : zhipu
+# @Time         : 2025/2/19 20:42
+# @Author       : betterme
+# @WeChat       : meutils
+# @Software     : PyCharm
+# @Description  : 
+
+from meutils.pipe import *
+from meutils.str_utils.json_utils import json_path
+from meutils.llm.clients import zhipuai_sdk_client, zhipuai_client
+
+from meutils.llm.openai_utils import to_openai_params
+from meutils.schemas.openai_types import chat_completion, chat_completion_chunk, CompletionRequest, \
+    ChatCompletionRequest
+
+
+class Completions(object):
+
+    def __init__(self, api_key: Optional[str] = None):
+        self.api_key = api_key
+
+    async def create(self, request: ChatCompletionRequest, search_result2md: bool = False):
+        pass
+
+    # async def _create(self, request: ChatCompletionRequest, search_result2md: bool = False):
+    #     request.stream = False
+    #     request.model = "web-search-pro"
+    #     request.messages = [{
+    #         "role": "user",
+    #         "content": request.last_content,
+    #     }]
+    #     data = to_openai_params(request)
+    #
+    #     search_completion = await zhipuai_client.chat.completions.create(**data)
+    #     logger.debug(search_completion.model_dump_json(indent=4))
+    #
+    #     if results := json_path(search_completion, '$..[keywords,query,search_result]'):
+    #         data = dict(zip(["keywords", "query", "search_result"], results))
+    #         if search_result2md:
+    #             global df
+    #
+    #             df = pd.DataFrame(data["search_result"])
+    #
+    #             df['title'] = [f"[{k}]({v})" for k, v in zip(df['title'], df['link'])]
+    #             df['media'] = [f"![{k}]({v})" for k, v in zip(df['media'], df['icon'])]
+    #
+    #             df = df[['title', 'media']]
+    #             df.index += 1
+    #             # {df_.to_markdown(index=False).replace('|:-', '|-').replace('-:|', '-|')}
+    #             data["search_result"] = df.to_markdown()
+    #         return data
+
+    async def search(self, q: str):
+        zhipuai_sdk_client.assistant.conversation(
+            model="web-search-pro",
+        )
+
+        # {
+        #     "role": "user",
+        #     "content": search_completion.model_dump_json(indent=4),
+        # }
+
+
+if __name__ == '__main__':
+    model = "web-search-pro"
+    # model = "tencent-search"
+
+    request = ChatCompletionRequest(
+        # model="baichuan4-turbo",
+        # model="xx",
+        # model="deepseek-r1",
+        # model="deepseek-r1:1.5b",
+        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"}
+        ],
+
+        stream=True
+    )
+    arun(Completions().create(request, search_result2md=True))

meutils/apis/search/metaso.py

@@ -13,7 +13,7 @@ from meutils.schemas.metaso_types import FEISHU_URL, BASE_URL, MetasoRequest, Me
 from meutils.decorators.retry import retrying
 from meutils.config_utils.lark_utils import get_next_token_for_polling
 from meutils.apis.proxy.ips import FEISHU_URL_METASO, get_one_proxy, get_proxies
-from meutils.schemas.openai_types import ChatCompletionRequest
+from meutils.schemas.openai_types import ChatCompletionRequest, CompletionRequest
 from meutils.notice.feishu import send_message
 
 from urllib.parse import quote_plus
@@ -95,7 +95,14 @@ async def get_access_token(session_id: Optional[str] = None):
         return tokens and tokens[0]
 
 
-async def create(request: ChatCompletionRequest):
+async def create(request: Union[ChatCompletionRequest, CompletionRequest]):
+    if isinstance(request, CompletionRequest):
+        request = ChatCompletionRequest(**request.model_dump())
+
+    if request.last_content == 'ping':
+        yield "pong"
+        return
+
     system_fingerprint = request.system_fingerprint
 
     engine_type = ''

meutils/apis/search/web_search.py

@@ -0,0 +1,132 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+# @Project      : AI.  @by PyCharm
+# @File         : web_search
+# @Time         : 2025/3/18 20:15
+# @Author       : betterme
+# @WeChat       : meutils
+# @Software     : PyCharm
+# @Description  : 
+
+from aiostream import stream
+from meutils.pipe import *
+from meutils.async_utils import sync_to_async
+
+from meutils.llm.clients import zhipuai_sdk_client
+from meutils.schemas.openai_types import chat_completion, chat_completion_chunk, CompletionRequest, ChatCompletionChunk
+
+
+def convert_citations(text):
+    """
+    # 示例使用
+    text = "这是一段包含【1†source】和【2†source】的文本"
+    result = convert_citations(text)
+    print(result)  # 输出: 这是一段包含[^1]和[^2]的文本
+    :param text:
+    :return:
+    """
+    # 匹配【数字†source】格式的引用
+    pattern = r'【(\d+)†source】'
+
+    # 替换为[^数字]格式
+    converted = re.sub(pattern, r'[^\1]', text)
+
+    return converted
+
+
+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 _
+
+    async def search(self, q: str):
+        return list(self._create(q))
+
+    def _create(self, request: Union[CompletionRequest, str]):
+        chunks = zhipuai_sdk_client.assistant.conversation(
+
+            assistant_id="659e54b1b8006379b4b2abd6",  # 搜索智能体
+            conversation_id=None,
+            model="glm-4-assistant",  # assistant-ppt
+            messages=[
+                {
+                    "role": "user",
+                    "content": [{
+                        "type": "text",
+                        # "text": "北京未来七天气温，做个折线图",
+                        # "text": "画条狗"
+                        "text": request.last_user_content if isinstance(request, CompletionRequest) else request,
+
+                    }]
+                }
+            ],
+            stream=True,
+            attachments=None,
+            metadata=None
+        )
+
+        references = []
+        buffer = []
+        for chunk in chunks:
+            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
+                continue
+
+            if isinstance(request, CompletionRequest):
+                if references:
+                    urls = [f"[^{i}]: [{ref['title']}]({ref['link']})\n" for i, ref in enumerate(references, 1)]
+                    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
+
+            else:
+                yield references
+                break
+
+
+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="moonshot-v1-8k",
+        # model="doubao",
+
+        messages=[
+            {"role": "user", "content": "《哪吒之魔童闹海》现在的票房是多少"}
+        ],
+
+        stream=True
+    )
+    # arun(Completions().search('周杰伦'))
+    arun(Completions().create(request))

meutils/apis/siliconflow/image_to_image.py

@@ -23,7 +23,7 @@ BASE_URL = "https://cloud.siliconflow.cn"
 FEISHU_URL = "https://xchatllm.feishu.cn/sheets/Bmjtst2f6hfMqFttbhLcdfRJnNf?sheet=InxiCF"
 FEISHU_URL_TOKEN = "https://xchatllm.feishu.cn/sheets/Bmjtst2f6hfMqFttbhLcdfRJnNf?sheet=xlvlrH"
 
-url = "https://api.siliconflow.com/v1/TencentARC/PhotoMaker/image-to-image"
+url = "https://api.siliconflow.cn/v1/TencentARC/PhotoMaker/image-to-image"
 
 PHOTOMAKER_STYLES = {
     "Photographic (Default)",
@@ -83,8 +83,8 @@ async def create(request: ImageRequest, api_key: Optional[str] = None):  # SD3
     headers = {
         "Authorization": f"Bearer {api_key}",
     }
-    base_url = "https://api.siliconflow.com/v1"
-    url = "https://api.siliconflow.com/v1/TencentARC/PhotoMaker/image-to-image"
+    base_url = "https://api.siliconflow.cn/v1"
+    url = "https://api.siliconflow.cn/v1/TencentARC/PhotoMaker/image-to-image"
 
     async with httpx.AsyncClient(base_url=base_url, headers=headers, timeout=100) as client:
         response = await client.post(f"/{request.model}/image-to-image", json=payload)

meutils/apis/siliconflow/images.py

@@ -75,7 +75,7 @@ async def generate(request: ImageRequest, api_key: Optional[str] = None):
         # {'model': 'flux-schnell', 'messages': [{'role': 'user', 'content': '写一个10个字的冷笑话'}]}
         return ImagesResponse(**request.model_dump())
 
-    if not request.model.startswith(("flux",)):  # 自动翻译
+    if not request.model.startswith(("flux", "black-forest-labs")):  # 自动翻译
         request.prompt = (
             await deeplx.translate(DeeplxRequest(text=request.prompt, target_lang="EN"))
         ).get("data", request.prompt)
@@ -161,6 +161,8 @@ if __name__ == '__main__':
 
     # request = FluxImageRequest(model="flux", prompt="a dog", size="1024x1024", num_inference_steps=1)
     # request = FluxImageRequest(model="flux-pro", prompt="a dog", size="10x10", num_inference_steps=1)
+
+    data = {'model': 'flux-schnell', 'prompt': '画一个2025年电脑如何一键重装系统win10教程详解的封面图', 'n': 1, 'size': '680x400'}
     request = FluxImageRequest(**data)
 
     print(request)
@@ -176,4 +178,4 @@ if __name__ == '__main__':
 
     arun(generate(request))
 
-    # https://api.siliconflow.com/v1
+    # https://api.siliconflow.cn/v1

meutils/apis/siliconflow/text_to_image.py

@@ -17,7 +17,7 @@ from meutils.schemas.translator_types import DeeplxRequest
 from meutils.decorators.retry import retrying
 from meutils.schemas.image_types import ASPECT_RATIOS
 
-BASE_URL = "https://api.siliconflow.com/v1"
+BASE_URL = "https://api.siliconflow.cn/v1"
 FEISHU_URL = "https://xchatllm.feishu.cn/sheets/Bmjtst2f6hfMqFttbhLcdfRJnNf?sheet=InxiCF"
 FEISHU_URL_TOKEN = "https://xchatllm.feishu.cn/sheets/Bmjtst2f6hfMqFttbhLcdfRJnNf?sheet=xlvlrH"
 

meutils/apis/siliconflow/utils.py

@@ -12,7 +12,7 @@ import os
 from meutils.pipe import *
 from meutils.decorators.retry import retrying
 
-url = "https://api.siliconflow.com/v1/user/info"
+url = "https://api.siliconflow.cn/v1/user/info"
 
 
 # todo: 付费不付费模型优化

meutils/config_utils/lark_utils/common.py

@@ -9,10 +9,13 @@
 # @Description  : httpx重试 transport = httpx.AsyncHTTPTransport(retries=3) # response.raise_for_status()
 
 from meutils.pipe import *
+from meutils.caches import rcache
+
 from meutils.decorators.retry import retrying
 from meutils.decorators.contextmanagers import try_catcher
 from meutils.notice.feishu import send_message
 from meutils.db.redis_db import redis_client, redis_aclient
+from typing import Optional, Union
 
 from urllib.parse import urlparse, parse_qs
 
@@ -73,6 +76,7 @@ def get_spreadsheet_values(
 
 
 @alru_cache(ttl=600)
+@rcache(ttl=600, serializer='pickle')  # 缓存
 async def aget_spreadsheet_values(
         spreadsheet_token: Optional[str] = None,
         sheet_id: Optional[str] = None,
@@ -381,5 +385,5 @@ if __name__ == '__main__':
     # with timer():
     #     token = arun(get_next_token(feishu_url))
 
-    # FEISHU_URL = "https://xchatllm.feishu.cn/sheets/XfjqszII3hZAEvtTOgecOgv2nye?sheet=c14b34"
-    # print(arun(get_series(FEISHU_URL)))
+    FEISHU_URL = "https://xchatllm.feishu.cn/sheets/XfjqszII3hZAEvtTOgecOgv2nye?sheet=c14b34"
+    print(arun(get_series(FEISHU_URL)))

meutils/data/VERSION

@@ -1 +1 @@
-2025.03.14.08.43.03
+2025.03.19.19.13.35

meutils/data/oneapi/index.html

@@ -235,6 +235,15 @@
                     },
                 ],
                 changeLogs: [
+                    {
+                        date: '2025.03.15',
+                        title: '🔥🔥🔥 上线新模型',
+                        content: [
+                            'suno风控成本上涨，单价略微上涨，明日生效',
+                            'google/gemma-3-27b-it：Gemma 3 引入了多模态性，支持视觉语言输入和文本输出。它可以处理多达 128k 个标记的上下文窗口，理解 140 多种语言，并提供改进的数学、推理和聊天功能，包括结构化输出和函数调用。Gemma 3 27B 是 Google 最新的开源模型，是 Gemma 2 的继任者',
+
+                        ],
+                    },
                     {
                         date: '2025.03.06',
                         title: '🔥🔥🔥 上线新模型',

meutils/io/files_utils.py

@@ -224,7 +224,18 @@ if __name__ == '__main__':
 
     file = Path("/Users/betterme/PycharmProjects/AI/MeUtils/meutils/io/x.py").read_bytes()
 
-    arun(to_url_fal([file] * 1))
+    # arun(to_url_fal([file] * 1))
+
+    print(mimetypes.guess_type("http://url"))
+    print(mimetypes.guess_type("http://url.pdf"))
+    print(mimetypes.guess_type("http://url.php"))
+
+    print(mimetypes.guess_type("xx.txt"))
+    print(mimetypes.guess_type("xx.html"))
+    print(mimetypes.guess_type("xx.mp3"))
+    print(mimetypes.guess_type("xx.mp4"))
+
+    # print(guess_mime_type("http://url")) # application # msword
 
     # arun(to_bytes(None))
 

meutils/io/openai_files.py

@@ -17,6 +17,7 @@ from meutils.io.files_utils import to_bytes, guess_mime_type
 from meutils.llm.clients import moonshot_client, zhipuai_client, APIStatusError
 from meutils.notice.feishu import send_message as _send_message, FILES
 from meutils.caches import cache, rcache
+from meutils.apis.jina import url_reader
 
 # from openai.types.file_object import FileObject
 
@@ -52,17 +53,28 @@ async def delete_files(client, threshold: int = 666):
         await asyncio.gather(*tasks)
 
 
-@rcache(ttl=1 * 3600)
+@rcache(ttl=7 * 24 * 3600)
 async def _file_extract(file):
     """
 
     :param file: url bytes path
     :return:
     """
+    # url
 
     filename = Path(file).name if isinstance(file, str) else 'untitled'
     mime_type = guess_mime_type(file)
 
+    if str(file).startswith("http") and mime_type in {"application/octet-stream", "text/html"}:
+        logger.debug(f"jina reader")
+        markdown_content = await url_reader(file)
+        return {
+            'filename': filename,
+            'type': 'file',
+            'file_type': "text/plain",
+            'content': markdown_content,
+        }
+
     file: bytes = await to_bytes(file)
 
     for i, client in enumerate([moonshot_client, zhipuai_client]):
@@ -134,5 +146,18 @@ if __name__ == '__main__':
     # with timer():
     #     r = arun(file_extract(file, provider='kimi'))
 
+    # with timer():
+    #     arun(file_extract(file))
+
+    # with timer():
+    #     arun(file_extract("https://top.baidu.com/board?tab=realtime"))
+
     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"
+
         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"))
+        # arun(file_extract("/Users/betterme/PycharmProjects/AI/data/098【采集】基于室内定位导航的医院解决方案.pdf"))
+        # arun(file_extract("//Users/betterme/PycharmProjects/AI/data/《 纺织行业场景化解决方案-客户介绍材料》.pptx"))