Delete agent_for_ui/chatgpt.py
Browse files- agent_for_ui/chatgpt.py +0 -168
agent_for_ui/chatgpt.py
DELETED
@@ -1,168 +0,0 @@
|
|
1 |
-
import requests
|
2 |
-
import os
|
3 |
-
import openai
|
4 |
-
import backoff
|
5 |
-
import requests
|
6 |
-
import time
|
7 |
-
import json
|
8 |
-
completion_tokens = prompt_tokens = 0
|
9 |
-
|
10 |
-
# api_key = "sk-cTJqhJUkALz7qQbQ7a027280A1Df499883A2116bD8F1765d"
|
11 |
-
# if api_key != "":
|
12 |
-
# openai.api_key = api_key
|
13 |
-
# else:
|
14 |
-
# print("Warning: OPENAI_API_KEY is not set")
|
15 |
-
#
|
16 |
-
# api_base = os.getenv("OPENAI_API_BASE", "https://api.aiguoguo199.com/v1")
|
17 |
-
# if api_base != "":
|
18 |
-
# print("Warning: OPENAI_API_BASE is set to {}".format(api_base))
|
19 |
-
# openai.api_base = api_base
|
20 |
-
#
|
21 |
-
#
|
22 |
-
# # @backoff.on_exception(backoff.expo)
|
23 |
-
# # def completions_with_backoff(**kwargs):
|
24 |
-
# # return openai.ChatCompletion.create(**kwargs)
|
25 |
-
# def chatgpt(prompt):
|
26 |
-
# """
|
27 |
-
#
|
28 |
-
# """
|
29 |
-
# response = openai.ChatCompletion.create(
|
30 |
-
# model="gpt-3.5-turbo",
|
31 |
-
# messages=[
|
32 |
-
# # {'role': 'system', 'content': persona},
|
33 |
-
# {'role': 'user', 'content': prompt}
|
34 |
-
# ],
|
35 |
-
# temperature=0.1,
|
36 |
-
# max_tokens=1000,
|
37 |
-
# n=1,
|
38 |
-
# stop=None,
|
39 |
-
# top_p=0.95
|
40 |
-
# )
|
41 |
-
# print("response: ", response)
|
42 |
-
# return response
|
43 |
-
|
44 |
-
# def gpt(prompt, model="gpt-3.5-turbo", temperature=0.7, max_tokens=1000, n=1, stop=None) -> list:
|
45 |
-
# # 使用GPT模型生成对话,返回生成的消息列表
|
46 |
-
# messages = [{"role": "user", "content": prompt}]
|
47 |
-
# return chatgpt(messages, model=model, temperature=temperature, max_tokens=max_tokens, n=n, stop=stop)
|
48 |
-
|
49 |
-
|
50 |
-
# def chatgpt(messages, model="gpt-3.5-turbo", temperature=0.7, max_tokens=1000, n=1, stop=None) -> list:
|
51 |
-
# global completion_tokens, prompt_tokens
|
52 |
-
# outputs = []
|
53 |
-
# while n > 0:
|
54 |
-
# # 限制每次请求的数量,以避免出现请求过多的情况
|
55 |
-
# cnt = min(n, 20)
|
56 |
-
# n -= cnt
|
57 |
-
# # 使用backoff算法重试生成对话请求,返回生成的消息列表
|
58 |
-
# res = completions_with_backoff(model=model, messages=messages, temperature=temperature, max_tokens=max_tokens,
|
59 |
-
# n=cnt, stop=stop)
|
60 |
-
# '''
|
61 |
-
# res是一个字典,包含从GPT模型返回的响应信息。res["choices"]是一个列表,其中包含GPT模型生成的若干个响应。每个响应是一个字典,包含以下键:
|
62 |
-
# "text":一个字符串,表示GPT模型生成的响应文本。
|
63 |
-
# "finish_reason":一个字符串,表示GPT模型停止生成响应的原因。
|
64 |
-
# "index":一个整数,表示GPT模型生成响应的索引。
|
65 |
-
# "logprobs":一个字典,表示GPT模型生成响应时的对数概率。
|
66 |
-
# "tokens":一个列表,表示GPT模型生成响应时使用的令牌(标记)序列。
|
67 |
-
# "message":一个字典,包含以下键:
|
68 |
-
# "content":一个字符串,表示GPT模型生成的响应文本。
|
69 |
-
# "metadata":一个字典,包含与响应相关的元数据。
|
70 |
-
# '''
|
71 |
-
# outputs.extend([choice["message"]["content"] for choice in res["choices"]])
|
72 |
-
# # log completion tokens
|
73 |
-
# # 记录使用的生成响应和接受提示使用的token数。
|
74 |
-
# completion_tokens += res["usage"]["completion_tokens"]
|
75 |
-
# prompt_tokens += res["usage"]["prompt_tokens"]
|
76 |
-
# return outputs
|
77 |
-
|
78 |
-
# ####################### Xiaomi API接口 ##########################
|
79 |
-
# def chatgpt(prompt, model="gpt-4-turbo-2024-04-09", temperature=0.1, max_tokens=1000, n=1, stop=None) -> list:
|
80 |
-
# headers = {'Content-Type': 'application/json'}
|
81 |
-
# url = 'http://10.221.105.108:19005' # gpt-4/3.5 for ChatCompletion
|
82 |
-
# outputs = []
|
83 |
-
# cnt = min(n, 20)
|
84 |
-
# while n > 0:
|
85 |
-
# if model == "gpt-4-turbo-2024-04-09":
|
86 |
-
# data = {"uid": "xx9ty", "prompt": prompt, "history": [], "model": "gpt-4-0314",
|
87 |
-
# "max_tokens": max_tokens, "n": cnt, "temperature": temperature, "stop": stop} # if n>1, response of data is a List.
|
88 |
-
# elif model == "gpt-3.5-turbo":
|
89 |
-
# data = {"uid": "xx9ty", "prompt": prompt, "history": [],
|
90 |
-
# "max_tokens": max_tokens, "n": cnt, "temperature": temperature, "stop": stop}
|
91 |
-
# response = requests.post(url, json=data, headers=headers)
|
92 |
-
# try:
|
93 |
-
# res = response.json()
|
94 |
-
# if isinstance(res['response'], list):
|
95 |
-
# assert cnt > 1
|
96 |
-
# outputs.extend(res['response'])
|
97 |
-
# elif isinstance(res['response'], str):
|
98 |
-
# assert cnt == 1
|
99 |
-
# outputs.append(res['response'])
|
100 |
-
# else:
|
101 |
-
# print("None of outputs!")
|
102 |
-
# n -= cnt
|
103 |
-
# cnt = min(n, 20)
|
104 |
-
# except Exception as e:
|
105 |
-
# # print(e)
|
106 |
-
# pass
|
107 |
-
# # print(len(outputs))
|
108 |
-
# # for i, output in enumerate(outputs):
|
109 |
-
# # print(i, output)
|
110 |
-
# # print("data", data)
|
111 |
-
# return outputs
|
112 |
-
|
113 |
-
def get_model_response(self, prompt, images=None):
|
114 |
-
if images is None:
|
115 |
-
images = []
|
116 |
-
url = f"http://preview-general-llm.api.ai.srv/api/{self.model}/{self.user_name}"
|
117 |
-
if self.model == 'gpt-4-turbo-2024-04-09':
|
118 |
-
content = [
|
119 |
-
{
|
120 |
-
"type": 'text',
|
121 |
-
"text": prompt
|
122 |
-
}
|
123 |
-
]
|
124 |
-
messages = [
|
125 |
-
{
|
126 |
-
"role": 'user',
|
127 |
-
"content": content
|
128 |
-
}
|
129 |
-
]
|
130 |
-
else:
|
131 |
-
print(f"Unsupported Model: {self.model}")
|
132 |
-
assert False
|
133 |
-
try_cnt = 0
|
134 |
-
while try_cnt < 5:
|
135 |
-
try_cnt += 1
|
136 |
-
resp = requests.post(
|
137 |
-
url=url,
|
138 |
-
json={
|
139 |
-
"messages": messages,
|
140 |
-
"max_tokens": 1000,
|
141 |
-
"temperature": 0.1,
|
142 |
-
"seed": 1234,
|
143 |
-
# "model": self.model,
|
144 |
-
# "user_name": self.user_name
|
145 |
-
})
|
146 |
-
try:
|
147 |
-
res = resp.json()
|
148 |
-
if 'Error code' in res['response']:
|
149 |
-
print("[API ERROR]!!! API Request Error!!!")
|
150 |
-
# print_with_color(json.dumps(res), "red")
|
151 |
-
time.sleep(20)
|
152 |
-
continue
|
153 |
-
return res['response']
|
154 |
-
except Exception as e:
|
155 |
-
print("[API ERROR]!!! Json Decode Error!!!")
|
156 |
-
# print_with_color(resp.text, "red")
|
157 |
-
time.sleep(20)
|
158 |
-
continue
|
159 |
-
return "API ERROR"
|
160 |
-
|
161 |
-
def gpt_usage(backend="gpt-4"):
|
162 |
-
# 根据使用的GPT模型计算使用的completion_tokens和prompt_tokens,并计算出费用
|
163 |
-
global completion_tokens, prompt_tokens
|
164 |
-
if backend == "gpt-4":
|
165 |
-
cost = completion_tokens / 1000 * 0.06 + prompt_tokens / 1000 * 0.03
|
166 |
-
elif backend == "gpt-3.5-turbo":
|
167 |
-
cost = (completion_tokens + prompt_tokens) / 1000 * 0.0002
|
168 |
-
return {"completion_tokens": completion_tokens, "prompt_tokens": prompt_tokens, "cost": cost}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|