-
Notifications
You must be signed in to change notification settings - Fork 1
/
openai_wrapper.py
230 lines (208 loc) · 9.04 KB
/
openai_wrapper.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
import os
import openai
from openai import OpenAI
from tenacity import (
retry,
stop_after_attempt,
wait_random_exponential,
) # for exponential backoff
import logging
import json
import base64
from time import sleep
# Configure logging
logging.basicConfig(level=logging.INFO)
# Error callback function
def log_retry_error(retry_state):
logging.error(f"Retrying due to error: {retry_state.outcome.exception()}")
DEFAULT_CONFIG = {
"model": "gpt-4-turbo-preview",
"temperature": 0.0,
"max_tokens": 2048,
"top_p": 1.0,
"frequency_penalty": 0.0,
"presence_penalty": 0.0,
"stop": None
}
class OpenAIWrapper:
def __init__(self, config = DEFAULT_CONFIG, system_message="You are a helpful assistant."):
openai.api_key = os.environ.get("OPENAI_API_KEY")
print("set api key:", os.environ["OPENAI_API_KEY"])
if os.environ.get("OPENAI_ORG_KEY") is not None:
print("set organization key:", os.environ["OPENAI_ORG_KEY"])
openai.organization = os.environ.get("OPENAI_ORG_KEY")
# if os.environ.get("USE_AZURE")=="True":
# print("using azure api")
# openai.api_type = "azure"
# openai.api_base = os.environ.get("API_BASE")
# openai.api_version = os.environ.get("API_VERSION")
self.config = config
print("api config:", config, '\n')
self.client = OpenAI()
# count total tokens
self.completion_tokens = 0
self.prompt_tokens = 0
# system message
self.system_message = system_message # "You are an AI assistant that helps people find information."
# retry using tenacity
@retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(3), retry_error_callback=log_retry_error)
def completions_with_backoff(self, **kwargs):
# print("making api call:", kwargs)
# print("====================================")
# return openai.ChatCompletion.create(**kwargs)
return self.client.chat.completions.create(**kwargs)
@retry(wait=wait_random_exponential(min=1, max=60), stop=stop_after_attempt(3), retry_error_callback=log_retry_error)
def assistant_w_code_interpreter_with_backoff(self, **kwargs):
assistant = self.client.beta.assistants.create(
name = self.config["assistant_name"],
instructions = self.config["assistant_instruction"],
tools=[{"type": "code_interpreter"}],
model=self.config["model"]
)
thread = self.client.beta.threads.create()
message = self.client.beta.threads.messages.create(
thread_id=thread.id,
role="user",
content=kwargs["prompt"]
)
print("assistant message:", message)
run = self.client.beta.threads.runs.create(
thread_id=thread.id,
assistant_id=assistant.id
)
max_tries = 50
while True:
retrieved_run = self.client.beta.threads.runs.retrieve(
thread_id=thread.id,
run_id=run.id
)
if retrieved_run.status == "completed":
break
print(f"waiting for completion on run_id {run.id}...")
sleep(10)
max_tries -= 1
if max_tries == 0:
return None, None
messages = self.client.beta.threads.messages.list(
thread_id=thread.id
)
run_steps = self.client.beta.threads.runs.steps.list(
thread_id=thread.id,
run_id=run.id
)
return messages, run_steps
# Function to encode the image
def encode_image(self, image_path):
with open(image_path, "rb") as image_file:
return base64.b64encode(image_file.read()).decode('utf-8')
def get_input_message(self, text, images=[]):
if images == []:
return {"role":"user", "content":text}
else:
mes = {
"role":"user",
"content": [
{"type": "text", "text": text}
]
}
print("adding images:", images)
for img_path in images:
img = self.encode_image(img_path)
mes["content"].append({
"type": "image_url",
"image_url": {
"url": f"data:image/jpeg;base64,{img}"
# "url": "https://upload.wikimedia.org/wikipedia/commons/thumb/d/dd/Gfp-wisconsin-madison-the-nature-boardwalk.jpg/2560px-Gfp-wisconsin-madison-the-nature-boardwalk.jpg",
}
})
return mes
def run(self, prompt, n=1, system_message="", images=[]):
"""
prompt: str
n: int, total number of generations specified
"""
try:
# overload system message
if system_message != "":
sys_m = system_message
else:
sys_m = self.system_message
user_mes = self.get_input_message(prompt, images)
if sys_m != "":
# print("adding system message:", sys_m)
messages = [
{"role":"system", "content":sys_m},
user_mes
]
else:
messages = [user_mes]
print("input messages:", messages, "\n")
text_outputs = []
raw_responses = []
rets = []
while n > 0:
cnt = min(n, 10) # number of generations per api call
n -= cnt
# print("messages:", messages, self.config)
if self.config["api_type"] == "chat_completion":
chat_configs = {key: value for key, value in self.config.items() if key != "api_type"}
res = self.completions_with_backoff(messages=messages, n=cnt, **chat_configs)
text_outputs.extend([ {
"content": choice.message.content,
"role": choice.message.role,
"function_call": getattr(choice.message, "function_call", None),
"tool_calls": getattr(choice.message, "tool_calls", None),
"finish_reason": getattr(choice, "finish_reason", None),
"index": getattr(choice, "index", None),
} for choice in res.choices])
# add prompt to log
ret = {
"id": res.id,
"model": res.model,
"prompt": prompt,
"system_message": sys_m,
"text_outputs": text_outputs
}
rets.append(ret)
raw_responses.append(res)
# log completion tokens
self.completion_tokens += res.usage.completion_tokens
self.prompt_tokens += res.usage.prompt_tokens
elif self.config["api_type"] == "assistant_code_interpreter":
res_messages, run_steps = self.assistant_w_code_interpreter_with_backoff(prompt=prompt)
tool_calls = []
if run_steps is not None:
for run_step in run_steps.data:
if hasattr(run_step.step_details, "tool_calls"):
for tool_call in run_step.step_details.tool_calls:
tool_calls.append(
{
"tool_type": "code_interpreter",
"input": tool_call.code_interpreter.input,
"output": tool_call.code_interpreter.outputs[0].logs
}
)
res = {
"messages": [{"content":res_messages.data[i].content[0].text.value, "role":res_messages.data[i].role} for i in range(len(res_messages.data))],
"tool_calls":tool_calls
}
ret = res['messages'][0]
rets.append(ret)
raw_responses.append(res)
else:
return [], []
if len(rets) == 1:
rets = rets[0]
return rets, raw_responses
except Exception as e:
print("an error occurred:", e)
return [], []
def compute_gpt_usage(self):
model = self.config["model"]
if model == "gpt-4-1106-preview":
cost = self.completion_tokens / 1000 * 0.01 + self.prompt_tokens / 1000 * 0.03
elif model == "gpt-3.5-turbo-1106":
cost = self.completion_tokens / 1000 * 0.001 + self.prompt_tokens / 1000 * 0.002
else:
cost = 0 # TODO: add custom cost calculation for other engines
return {"completion_tokens": self.completion_tokens, "prompt_tokens": self.prompt_tokens, "cost": cost}