Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -546,9 +546,197 @@ def check_tokens():
|
|
546 |
|
547 |
return jsonify(results)
|
548 |
|
549 |
-
|
550 |
-
|
551 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
552 |
return jsonify({"error": "Unauthorized"}), 401
|
553 |
|
554 |
data = request.get_json()
|
@@ -656,7 +844,8 @@ def handsome_chat_completions():
|
|
656 |
"index": 0,
|
657 |
"message": {
|
658 |
"role": "assistant",
|
659 |
-
"content":
|
|
|
660 |
},
|
661 |
"finish_reason": "length",
|
662 |
}
|
@@ -667,7 +856,9 @@ def handsome_chat_completions():
|
|
667 |
"total_tokens": 0
|
668 |
}
|
669 |
}
|
670 |
-
|
|
|
|
|
671 |
except (KeyError, ValueError, IndexError) as e:
|
672 |
logging.error(
|
673 |
f"解析响应 JSON 失败: {e}, "
|
@@ -685,6 +876,7 @@ def handsome_chat_completions():
|
|
685 |
"message": {
|
686 |
"role": "assistant",
|
687 |
"content": "Failed to process image data",
|
|
|
688 |
},
|
689 |
"finish_reason": "stop",
|
690 |
}
|
@@ -702,10 +894,6 @@ def handsome_chat_completions():
|
|
702 |
f"使用的模型: {model_name}"
|
703 |
)
|
704 |
|
705 |
-
with data_lock:
|
706 |
-
request_timestamps.append(time.time())
|
707 |
-
token_counts.append(0) # Image generation doesn't use tokens
|
708 |
-
|
709 |
return jsonify(response_data)
|
710 |
except requests.exceptions.RequestException as e:
|
711 |
logging.error(f"请求转发异常: {e}")
|
@@ -718,191 +906,69 @@ def handsome_chat_completions():
|
|
718 |
TEST_MODEL_ENDPOINT,
|
719 |
headers=headers,
|
720 |
json=data,
|
721 |
-
stream=
|
722 |
timeout=60
|
723 |
)
|
724 |
if response.status_code == 429:
|
725 |
return jsonify(response.json()), 429
|
726 |
-
|
727 |
-
|
728 |
-
|
729 |
-
|
730 |
-
|
731 |
-
|
732 |
-
|
733 |
-
|
734 |
-
|
735 |
-
|
736 |
-
|
737 |
-
|
738 |
-
|
739 |
-
|
740 |
-
|
741 |
-
|
742 |
-
|
743 |
-
|
744 |
-
|
745 |
-
|
746 |
-
|
747 |
-
|
748 |
-
|
749 |
-
|
750 |
-
|
751 |
-
|
752 |
-
|
753 |
-
|
754 |
-
|
755 |
-
|
756 |
-
|
757 |
-
|
758 |
-
|
759 |
-
|
760 |
-
|
761 |
-
|
762 |
-
]["completion_tokens"]
|
763 |
-
|
764 |
-
if (
|
765 |
-
"choices" in response_json and
|
766 |
-
len(response_json["choices"]) > 0 and
|
767 |
-
"delta" in response_json["choices"][0] and
|
768 |
-
"content" in response_json[
|
769 |
-
"choices"
|
770 |
-
][0]["delta"]
|
771 |
-
):
|
772 |
-
response_content += response_json[
|
773 |
-
"choices"
|
774 |
-
][0]["delta"]["content"]
|
775 |
-
|
776 |
-
if (
|
777 |
-
"usage" in response_json and
|
778 |
-
"prompt_tokens" in response_json["usage"]
|
779 |
-
):
|
780 |
-
prompt_tokens = response_json[
|
781 |
-
"usage"
|
782 |
-
]["prompt_tokens"]
|
783 |
-
|
784 |
-
except (
|
785 |
-
KeyError,
|
786 |
-
ValueError,
|
787 |
-
IndexError
|
788 |
-
) as e:
|
789 |
-
logging.error(
|
790 |
-
f"解析流式响应单行 JSON 失败: {e}, "
|
791 |
-
f"行内容: {line}"
|
792 |
)
|
793 |
|
794 |
-
|
795 |
-
messages = data.get("messages", [])
|
796 |
-
for message in messages:
|
797 |
-
if message["role"] == "user":
|
798 |
-
if isinstance(message["content"], str):
|
799 |
-
user_content += message["content"] + " "
|
800 |
-
elif isinstance(message["content"], list):
|
801 |
-
for item in message["content"]:
|
802 |
-
if (
|
803 |
-
isinstance(item, dict) and
|
804 |
-
item.get("type") == "text"
|
805 |
-
):
|
806 |
-
user_content += (
|
807 |
-
item.get("text", "") +
|
808 |
-
" "
|
809 |
-
)
|
810 |
-
|
811 |
-
user_content = user_content.strip()
|
812 |
-
|
813 |
-
user_content_replaced = user_content.replace(
|
814 |
-
'\n', '\\n'
|
815 |
-
).replace('\r', '\\n')
|
816 |
-
response_content_replaced = response_content.replace(
|
817 |
-
'\n', '\\n'
|
818 |
-
).replace('\r', '\\n')
|
819 |
-
|
820 |
-
logging.info(
|
821 |
-
f"使用的key: {api_key}, "
|
822 |
-
f"提示token: {prompt_tokens}, "
|
823 |
-
f"输出token: {completion_tokens}, "
|
824 |
-
f"首字用时: {first_token_time:.4f}秒, "
|
825 |
-
f"总共用时: {total_time:.4f}秒, "
|
826 |
-
f"使用的模型: {model_name}, "
|
827 |
-
f"用户的内容: {user_content_replaced}, "
|
828 |
-
f"输出的内容: {response_content_replaced}"
|
829 |
-
)
|
830 |
-
|
831 |
-
with data_lock:
|
832 |
-
request_timestamps.append(time.time())
|
833 |
-
token_counts.append(prompt_tokens+completion_tokens)
|
834 |
|
835 |
-
|
836 |
-
|
837 |
-
|
838 |
-
|
839 |
-
|
840 |
-
|
841 |
-
end_time = time.time()
|
842 |
-
response_json = response.json()
|
843 |
-
total_time = end_time - start_time
|
844 |
|
845 |
-
|
846 |
-
|
847 |
-
|
848 |
-
|
849 |
-
|
850 |
-
|
851 |
-
|
852 |
-
|
853 |
-
|
854 |
-
|
855 |
-
|
856 |
-
f"完整内容: {response_json}"
|
857 |
-
)
|
858 |
-
prompt_tokens = 0
|
859 |
-
completion_tokens = 0
|
860 |
-
response_content = ""
|
861 |
-
|
862 |
-
user_content = ""
|
863 |
-
messages = data.get("messages", [])
|
864 |
-
for message in messages:
|
865 |
-
if message["role"] == "user":
|
866 |
-
if isinstance(message["content"], str):
|
867 |
-
user_content += message["content"] + " "
|
868 |
-
elif isinstance(message["content"], list):
|
869 |
-
for item in message["content"]:
|
870 |
-
if (
|
871 |
-
isinstance(item, dict) and
|
872 |
-
item.get("type") == "text"
|
873 |
-
):
|
874 |
-
user_content += (
|
875 |
-
item.get("text", "") +
|
876 |
-
" "
|
877 |
-
)
|
878 |
-
|
879 |
-
user_content = user_content.strip()
|
880 |
-
|
881 |
-
user_content_replaced = user_content.replace(
|
882 |
-
'\n', '\\n'
|
883 |
-
).replace('\r', '\\n')
|
884 |
-
response_content_replaced = response_content.replace(
|
885 |
-
'\n', '\\n'
|
886 |
-
).replace('\r', '\\n')
|
887 |
-
|
888 |
-
logging.info(
|
889 |
-
f"使用的key: {api_key}, "
|
890 |
-
f"提示token: {prompt_tokens}, "
|
891 |
-
f"输出token: {completion_tokens}, "
|
892 |
-
f"首字用时: 0, "
|
893 |
-
f"总共用时: {total_time:.4f}秒, "
|
894 |
-
f"使用的模型: {model_name}, "
|
895 |
-
f"用户的内容: {user_content_replaced}, "
|
896 |
-
f"输出的内容: {response_content_replaced}"
|
897 |
-
)
|
898 |
-
with data_lock:
|
899 |
-
request_timestamps.append(time.time())
|
900 |
-
if "prompt_tokens" in response_json["usage"] and "completion_tokens" in response_json["usage"]:
|
901 |
-
token_counts.append(response_json["usage"]["prompt_tokens"] + response_json["usage"]["completion_tokens"])
|
902 |
-
else:
|
903 |
-
token_counts.append(0)
|
904 |
-
|
905 |
-
return jsonify(response_json)
|
906 |
|
907 |
except requests.exceptions.RequestException as e:
|
908 |
logging.error(f"请求转发异常: {e}")
|
|
|
546 |
|
547 |
return jsonify(results)
|
548 |
|
549 |
+
import os
|
550 |
+
import time
|
551 |
+
import logging
|
552 |
+
import requests
|
553 |
+
import json
|
554 |
+
import uuid
|
555 |
+
from flask import Flask, request, jsonify
|
556 |
+
|
557 |
+
API_ENDPOINT = "https://api.siliconflow.cn/v1/user/info"
|
558 |
+
TEST_MODEL_ENDPOINT = "https://api.siliconflow.cn/v1/chat/completions"
|
559 |
+
EMBEDDINGS_ENDPOINT = "https://api.siliconflow.cn/v1/embeddings"
|
560 |
+
|
561 |
+
text_models = []
|
562 |
+
free_text_models = []
|
563 |
+
embedding_models = []
|
564 |
+
free_embedding_models = []
|
565 |
+
image_models = []
|
566 |
+
free_image_models = []
|
567 |
+
|
568 |
+
invalid_keys_global = []
|
569 |
+
free_keys_global = []
|
570 |
+
unverified_keys_global = []
|
571 |
+
valid_keys_global = []
|
572 |
+
model_key_indices = {}
|
573 |
+
|
574 |
+
request_timestamps = []
|
575 |
+
token_counts = []
|
576 |
+
data_lock = None
|
577 |
+
|
578 |
+
FREE_MODEL_TEST_KEY = (
|
579 |
+
"sk-bmjbjzleaqfgtqfzmcnsbagxrlohriadnxqrzfocbizaxukw"
|
580 |
+
)
|
581 |
+
|
582 |
+
FREE_IMAGE_LIST = [
|
583 |
+
"stabilityai/stable-diffusion-3-5-large",
|
584 |
+
"black-forest-labs/FLUX.1-schnell",
|
585 |
+
"stabilityai/stable-diffusion-3-medium",
|
586 |
+
"stabilityai/stable-diffusion-xl-base-1.0",
|
587 |
+
"stabilityai/stable-diffusion-2-1"
|
588 |
+
]
|
589 |
+
|
590 |
+
def get_credit_summary(api_key):
|
591 |
+
"""
|
592 |
+
使用 API 密钥获取额度信息。
|
593 |
+
"""
|
594 |
+
headers = {
|
595 |
+
"Authorization": f"Bearer {api_key}",
|
596 |
+
"Content-Type": "application/json"
|
597 |
+
}
|
598 |
+
try:
|
599 |
+
response = requests.get(API_ENDPOINT, headers=headers)
|
600 |
+
response.raise_for_status()
|
601 |
+
data = response.json().get("data", {})
|
602 |
+
total_balance = data.get("totalBalance", 0)
|
603 |
+
return {"total_balance": float(total_balance)}
|
604 |
+
except requests.exceptions.RequestException as e:
|
605 |
+
logging.error(f"获取额度信息失败,API Key:{api_key},错误信息:{e}")
|
606 |
+
return None
|
607 |
+
|
608 |
+
def test_model_availability(api_key, model_name):
|
609 |
+
"""
|
610 |
+
测试指定的模型是否可用。
|
611 |
+
"""
|
612 |
+
headers = {
|
613 |
+
"Authorization": f"Bearer {api_key}",
|
614 |
+
"Content-Type": "application/json"
|
615 |
+
}
|
616 |
+
try:
|
617 |
+
response = requests.post(
|
618 |
+
TEST_MODEL_ENDPOINT,
|
619 |
+
headers=headers,
|
620 |
+
json={
|
621 |
+
"model": model_name,
|
622 |
+
"messages": [{"role": "user", "content": "hi"}],
|
623 |
+
"max_tokens": 5,
|
624 |
+
"stream": False
|
625 |
+
},
|
626 |
+
timeout=5
|
627 |
+
)
|
628 |
+
if response.status_code == 429 or response.status_code == 200:
|
629 |
+
return True
|
630 |
+
else:
|
631 |
+
return False
|
632 |
+
except requests.exceptions.RequestException as e:
|
633 |
+
logging.error(
|
634 |
+
f"测试模型 {model_name} 可用性失败,"
|
635 |
+
f"API Key:{api_key},错误信息:{e}"
|
636 |
+
)
|
637 |
+
return False
|
638 |
+
|
639 |
+
def test_image_model_availability(api_key, model_name):
|
640 |
+
"""
|
641 |
+
测试指定的图像模型是否在 FREE_IMAGE_LIST 中。
|
642 |
+
如果在列表中,返回 True,否则返回 False。
|
643 |
+
"""
|
644 |
+
return model_name in FREE_IMAGE_LIST
|
645 |
+
|
646 |
+
def determine_request_type(model_name, model_list, free_model_list):
|
647 |
+
"""
|
648 |
+
根据用户请求的模型判断请求类型。
|
649 |
+
"""
|
650 |
+
if model_name in free_model_list:
|
651 |
+
return "free"
|
652 |
+
elif model_name in model_list:
|
653 |
+
return "paid"
|
654 |
+
else:
|
655 |
+
return "unknown"
|
656 |
+
|
657 |
+
def select_key(request_type, model_name):
|
658 |
+
"""
|
659 |
+
根据请求类型和模型名称选择合适的 KEY,
|
660 |
+
并实现轮询和重试机制。
|
661 |
+
"""
|
662 |
+
if request_type == "free":
|
663 |
+
available_keys = (
|
664 |
+
free_keys_global +
|
665 |
+
unverified_keys_global +
|
666 |
+
valid_keys_global
|
667 |
+
)
|
668 |
+
elif request_type == "paid":
|
669 |
+
available_keys = unverified_keys_global + valid_keys_global
|
670 |
+
else:
|
671 |
+
available_keys = (
|
672 |
+
free_keys_global +
|
673 |
+
unverified_keys_global +
|
674 |
+
valid_keys_global
|
675 |
+
)
|
676 |
+
|
677 |
+
if not available_keys:
|
678 |
+
return None
|
679 |
+
|
680 |
+
current_index = model_key_indices.get(model_name, 0)
|
681 |
+
|
682 |
+
for _ in range(len(available_keys)):
|
683 |
+
key = available_keys[current_index % len(available_keys)]
|
684 |
+
current_index += 1
|
685 |
+
|
686 |
+
if key_is_valid(key, request_type):
|
687 |
+
model_key_indices[model_name] = current_index
|
688 |
+
return key
|
689 |
+
else:
|
690 |
+
logging.warning(
|
691 |
+
f"KEY {key} 无效或达到限制,尝试下一个 KEY"
|
692 |
+
)
|
693 |
+
|
694 |
+
model_key_indices[model_name] = 0
|
695 |
+
return None
|
696 |
+
|
697 |
+
def key_is_valid(key, request_type):
|
698 |
+
"""
|
699 |
+
检查 KEY 是否有效,
|
700 |
+
根据不同的请求类型进行不同的检查。
|
701 |
+
"""
|
702 |
+
if request_type == "invalid":
|
703 |
+
return False
|
704 |
+
|
705 |
+
credit_summary = get_credit_summary(key)
|
706 |
+
if credit_summary is None:
|
707 |
+
return False
|
708 |
+
|
709 |
+
total_balance = credit_summary.get("total_balance", 0)
|
710 |
+
|
711 |
+
if request_type == "free":
|
712 |
+
return True
|
713 |
+
elif request_type == "paid" or request_type == "unverified":
|
714 |
+
return total_balance > 0
|
715 |
+
else:
|
716 |
+
return False
|
717 |
+
|
718 |
+
def check_authorization(request, authorization_key):
|
719 |
+
"""
|
720 |
+
检查请求头中的 Authorization 字段
|
721 |
+
是否匹配环境变量 AUTHORIZATION_KEY。
|
722 |
+
"""
|
723 |
+
if not authorization_key:
|
724 |
+
logging.warning("环境变量 AUTHORIZATION_KEY 未设置,请设置后重试。")
|
725 |
+
return False
|
726 |
+
|
727 |
+
auth_header = request.headers.get('Authorization')
|
728 |
+
if not auth_header:
|
729 |
+
logging.warning("请求头中缺少 Authorization 字段。")
|
730 |
+
return False
|
731 |
+
|
732 |
+
if auth_header != f"Bearer {authorization_key}":
|
733 |
+
logging.warning(f"无效的 Authorization 密钥:{auth_header}")
|
734 |
+
return False
|
735 |
+
|
736 |
+
return True
|
737 |
+
|
738 |
+
def handsome_chat_completions(request, authorization_key):
|
739 |
+
if not check_authorization(request, authorization_key):
|
740 |
return jsonify({"error": "Unauthorized"}), 401
|
741 |
|
742 |
data = request.get_json()
|
|
|
844 |
"index": 0,
|
845 |
"message": {
|
846 |
"role": "assistant",
|
847 |
+
"content": None,
|
848 |
+
"image_url": image_url if image_url else None, # Return URL in image_url
|
849 |
},
|
850 |
"finish_reason": "length",
|
851 |
}
|
|
|
856 |
"total_tokens": 0
|
857 |
}
|
858 |
}
|
859 |
+
if not image_url:
|
860 |
+
response_data["choices"][0]["message"]["content"] = "Failed to generate image"
|
861 |
+
|
862 |
except (KeyError, ValueError, IndexError) as e:
|
863 |
logging.error(
|
864 |
f"解析响应 JSON 失败: {e}, "
|
|
|
876 |
"message": {
|
877 |
"role": "assistant",
|
878 |
"content": "Failed to process image data",
|
879 |
+
"image_url": None,
|
880 |
},
|
881 |
"finish_reason": "stop",
|
882 |
}
|
|
|
894 |
f"使用的模型: {model_name}"
|
895 |
)
|
896 |
|
|
|
|
|
|
|
|
|
897 |
return jsonify(response_data)
|
898 |
except requests.exceptions.RequestException as e:
|
899 |
logging.error(f"请求转发异常: {e}")
|
|
|
906 |
TEST_MODEL_ENDPOINT,
|
907 |
headers=headers,
|
908 |
json=data,
|
909 |
+
stream=False,
|
910 |
timeout=60
|
911 |
)
|
912 |
if response.status_code == 429:
|
913 |
return jsonify(response.json()), 429
|
914 |
+
|
915 |
+
response.raise_for_status()
|
916 |
+
end_time = time.time()
|
917 |
+
response_json = response.json()
|
918 |
+
total_time = end_time - start_time
|
919 |
+
try:
|
920 |
+
prompt_tokens = response_json["usage"]["prompt_tokens"]
|
921 |
+
completion_tokens = response_json[
|
922 |
+
"usage"
|
923 |
+
]["completion_tokens"]
|
924 |
+
response_content = response_json[
|
925 |
+
"choices"
|
926 |
+
][0]["message"]["content"]
|
927 |
+
except (KeyError, ValueError, IndexError) as e:
|
928 |
+
logging.error(
|
929 |
+
f"解析非流式响应 JSON 失败: {e}, "
|
930 |
+
f"完整内容: {response_json}"
|
931 |
+
)
|
932 |
+
prompt_tokens = 0
|
933 |
+
completion_tokens = 0
|
934 |
+
response_content = ""
|
935 |
+
user_content = ""
|
936 |
+
messages = data.get("messages", [])
|
937 |
+
for message in messages:
|
938 |
+
if message["role"] == "user":
|
939 |
+
if isinstance(message["content"], str):
|
940 |
+
user_content += message["content"] + " "
|
941 |
+
elif isinstance(message["content"], list):
|
942 |
+
for item in message["content"]:
|
943 |
+
if (
|
944 |
+
isinstance(item, dict) and
|
945 |
+
item.get("type") == "text"
|
946 |
+
):
|
947 |
+
user_content += (
|
948 |
+
item.get("text", "") +
|
949 |
+
" "
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
950 |
)
|
951 |
|
952 |
+
user_content = user_content.strip()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
953 |
|
954 |
+
user_content_replaced = user_content.replace(
|
955 |
+
'\n', '\\n'
|
956 |
+
).replace('\r', '\\n')
|
957 |
+
response_content_replaced = response_content.replace(
|
958 |
+
'\n', '\\n'
|
959 |
+
).replace('\r', '\\n')
|
|
|
|
|
|
|
960 |
|
961 |
+
logging.info(
|
962 |
+
f"使用的key: {api_key}, "
|
963 |
+
f"提示token: {prompt_tokens}, "
|
964 |
+
f"输出token: {completion_tokens}, "
|
965 |
+
f"首字用时: 0, "
|
966 |
+
f"总共用时: {total_time:.4f}秒, "
|
967 |
+
f"使用的模型: {model_name}, "
|
968 |
+
f"用户的内容: {user_content_replaced}, "
|
969 |
+
f"输出的内容: {response_content_replaced}"
|
970 |
+
)
|
971 |
+
return jsonify(response_json)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
972 |
|
973 |
except requests.exceptions.RequestException as e:
|
974 |
logging.error(f"请求转发异常: {e}")
|