From 01a377d747a657b5c43d072a46ba5f39d9da0bd0 Mon Sep 17 00:00:00 2001 From: Your Name Date: Fri, 21 Apr 2023 17:37:48 +0800 Subject: [PATCH] =?UTF-8?q?=E8=BF=98=E5=8E=9FAPI=5FURL=E7=9A=84=E8=AE=BE?= =?UTF-8?q?=E7=BD=AE?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- config.py | 4 ++++ request_llm/bridge_all.py | 32 ++++++++++++++++++++++++++------ 2 files changed, 30 insertions(+), 6 deletions(-) diff --git a/config.py b/config.py index e24c68b..0ad723b 100644 --- a/config.py +++ b/config.py @@ -56,3 +56,7 @@ CONCURRENT_COUNT = 100 # 设置用户名和密码（不需要修改）（相关功能不稳定，与gradio版本和网络都相关，如果本地使用不建议加这个） # [("username", "password"), ("username2", "password2"), ...] AUTHENTICATION = [] + +# 重新URL重新定向，实现更换API_URL的作用（常规情况下，不要修改！！） +# 格式 {"https://api.openai.com/v1/chat/completions": "重定向的URL"} +API_URL_REDIRECT = {} diff --git a/request_llm/bridge_all.py b/request_llm/bridge_all.py index f1f4ee1..d1bc518 100644 --- a/request_llm/bridge_all.py +++ b/request_llm/bridge_all.py @@ -9,8 +9,9 @@ 2. predict_no_ui_long_connection：在实验过程中发现调用predict_no_ui处理长文档时，和openai的连接容易断掉，这个函数用stream的方式解决这个问题，同样支持多线程 """ import tiktoken -from functools import wraps, lru_cache +from functools import lru_cache from concurrent.futures import ThreadPoolExecutor +from toolbox import get_conf from .bridge_chatgpt import predict_no_ui_long_connection as chatgpt_noui from .bridge_chatgpt import predict as chatgpt_ui @@ -42,18 +43,37 @@ class LazyloadTiktoken(object): def decode(self, *args, **kwargs): encoder = self.get_encoder(self.model) return encoder.decode(*args, **kwargs) - + +# Endpoint 重定向 +API_URL_REDIRECT, = get_conf("API_URL_REDIRECT") +openai_endpoint = "https://api.openai.com/v1/chat/completions" +api2d_endpoint = "https://openai.api2d.net/v1/chat/completions" +# 兼容旧版的配置 +try: + API_URL, = get_conf("API_URL") + if API_URL != "https://api.openai.com/v1/chat/completions": + openai_endpoint = API_URL + print("警告！API_URL配置选项将被弃用，请更换为API_URL_REDIRECT配置") +except: + pass +# 新版配置 +if openai_endpoint in API_URL_REDIRECT: openai_endpoint = API_URL_REDIRECT[openai_endpoint] +if api2d_endpoint in API_URL_REDIRECT: api2d_endpoint = API_URL_REDIRECT[api2d_endpoint] + + +# 获取tokenizer tokenizer_gpt35 = LazyloadTiktoken("gpt-3.5-turbo") tokenizer_gpt4 = LazyloadTiktoken("gpt-4") get_token_num_gpt35 = lambda txt: len(tokenizer_gpt35.encode(txt, disallowed_special=())) get_token_num_gpt4 = lambda txt: len(tokenizer_gpt4.encode(txt, disallowed_special=())) + model_info = { # openai "gpt-3.5-turbo": { "fn_with_ui": chatgpt_ui, "fn_without_ui": chatgpt_noui, - "endpoint": "https://api.openai.com/v1/chat/completions", + "endpoint": openai_endpoint, "max_token": 4096, "tokenizer": tokenizer_gpt35, "token_cnt": get_token_num_gpt35, @@ -62,7 +82,7 @@ model_info = { "gpt-4": { "fn_with_ui": chatgpt_ui, "fn_without_ui": chatgpt_noui, - "endpoint": "https://api.openai.com/v1/chat/completions", + "endpoint": openai_endpoint, "max_token": 8192, "tokenizer": tokenizer_gpt4, "token_cnt": get_token_num_gpt4, @@ -72,7 +92,7 @@ model_info = { "api2d-gpt-3.5-turbo": { "fn_with_ui": chatgpt_ui, "fn_without_ui": chatgpt_noui, - "endpoint": "https://openai.api2d.net/v1/chat/completions", + "endpoint": api2d_endpoint, "max_token": 4096, "tokenizer": tokenizer_gpt35, "token_cnt": get_token_num_gpt35, @@ -81,7 +101,7 @@ model_info = { "api2d-gpt-4": { "fn_with_ui": chatgpt_ui, "fn_without_ui": chatgpt_noui, - "endpoint": "https://openai.api2d.net/v1/chat/completions", + "endpoint": api2d_endpoint, "max_token": 8192, "tokenizer": tokenizer_gpt4, "token_cnt": get_token_num_gpt4,