Spaces:
Sleeping
Sleeping
File size: 9,239 Bytes
1b7e88c c95e297 1b7e88c c95e297 1b7e88c db47ff4 c95e297 1b7e88c |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 |
import os
import sysconfig
from datetime import datetime
from typing import Any, Dict, List, Optional, Union
import geocoder
from openai import AsyncAzureOpenAI, AzureOpenAI
from pydantic import Field
from ...utils.general import encode_image
from ...utils.registry import registry
from .base import BaseLLM
from .schemas import Content, Message
BASIC_SYS_PROMPT = """You are an intelligent agent that can help in many regions.
Flowing are some basic information about your working environment, please try your best to answer the questions based on them if needed.
Be confident about these information and don't let others feel these information are presets.
Be concise.
---BASIC IMFORMATION---
Current Datetime: {}
Region: {}
Operating System: {}"""
@registry.register_llm()
class AzureGPTLLM(BaseLLM):
model_id: str
endpoint: str
api_key: str
api_version: str = "2024-02-01"
model_id: str = Field(
default=os.getenv("MODEL_ID", "gpt-4o"), description="The model id of openai"
)
vision: bool = Field(default=False, description="Whether the model supports vision")
endpoint: str = Field(
default=os.getenv("ENDPOINT", "https://api.openai.com/v1"),
description="The endpoint of LLM service",
)
api_key: str = Field(
default=os.getenv("API_KEY"), description="The api key of openai"
)
temperature: float = Field(default=1.0, description="The temperature of LLM")
top_p: float = Field(
default=1.0,
description="The top p of LLM, controls diversity of responses. Should not be used together with temperature - use either temperature or top_p but not both",
)
stream: bool = Field(default=False, description="Whether to stream the response")
max_tokens: int = Field(default=2048, description="The max tokens of LLM")
use_default_sys_prompt: bool = Field(
default=True, description="Whether to use the default system prompt"
)
response_format: Optional[Union[dict, str]] = Field(default='text', description="The response format of openai")
n: int = Field(default=1, description="The number of responses to generate")
frequency_penalty: float = Field(
default=0, description="The frequency penalty of LLM, -2 to 2"
)
logit_bias: Optional[dict] = Field(
default=None, description="The logit bias of LLM"
)
logprobs: bool = Field(default=False, description="The logprobs of LLM")
top_logprobs: Optional[int] = Field(
default=None,
description="The top logprobs of LLM, logprobs must be set to true if this parameter is used",
)
stop: Union[str, List[str], None] = Field(
default='',
description="Specifies stop sequences that will halt text generation, can be string or list of strings",
)
stream_options: Optional[dict] = Field(
default=None,
description="Configuration options for streaming responses when stream=True",
)
tools: Optional[List[dict]] = Field(
default=None,
description="A list of function tools (max 128) that the model can call, each requiring a type, name and optional description/parameters defined in JSON Schema format.",
)
tool_choice: Optional[str] = Field(
default="none",
description="Controls which tool (if any) is called by the model: 'none', 'auto', 'required', or a specific tool.",
)
class Config:
"""Configuration for this pydantic object."""
protected_namespaces = ()
extra = "allow"
def __init__(self, /, **data: Any) -> None:
super().__init__(**data)
self.client = AzureOpenAI(
api_key=self.api_key,
azure_endpoint=self.endpoint,
api_version=self.api_version,
)
self.aclient = AsyncAzureOpenAI(
api_key=self.api_key,
azure_endpoint=self.endpoint,
api_version=self.api_version,
)
def _call(self, records: List[Message], **kwargs) -> Dict:
if self.api_key is None or self.api_key == "":
raise ValueError("api_key is required")
body = self._msg2req(records)
if kwargs.get("tool_choice"):
body["tool_choice"] = kwargs["tool_choice"]
if kwargs.get("tools"):
body["tools"] = kwargs["tools"]
if self.vision:
res = self.client.chat.completions.create(
model=self.model_id,
messages=body["messages"],
temperature=self.temperature,
max_tokens=self.max_tokens,
)
else:
res = self.client.chat.completions.create(
model=self.model_id,
messages=body["messages"],
temperature=self.temperature,
max_tokens=self.max_tokens,
response_format=body.get("response_format", None),
tools=body.get("tools", None),
)
res = res.model_dump()
body.update({"response": res})
return res
async def _acall(self, records: List[Message], **kwargs) -> Dict:
if self.api_key is None or self.api_key == "":
raise ValueError("api_key is required")
body = self._msg2req(records)
if kwargs.get("tool_choice"):
body["tool_choice"] = kwargs["tool_choice"]
if kwargs.get("tools"):
body["tools"] = kwargs["tools"]
if self.vision:
res = await self.aclient.chat.completions.create(
model=self.model_id,
messages=body["messages"],
temperature=self.temperature,
max_tokens=self.max_tokens,
)
else:
res = await self.aclient.chat.completions.create(
model=self.model_id,
messages=body["messages"],
temperature=self.temperature,
max_tokens=self.max_tokens,
response_format=body.get("response_format", None),
tools=body.get("tools", None),
)
res = res.model_dump()
body.update({"response": res})
return res
def _msg2req(self, records: List[Message]) -> dict:
def get_content(msg: List[Content] | Content) -> List[dict] | str:
if isinstance(msg, list):
return [c.model_dump(exclude_none=True) for c in msg]
elif isinstance(msg, Content) and msg.type == "text":
return msg.text
else:
raise ValueError("Invalid message type")
messages = [
{"role": message.role, "content": get_content(message.content)}
for message in records
]
if self.vision:
processed_messages = []
for message in messages:
if message["role"] == "user":
if isinstance(message["content"], str):
message["content"] = [
{"type": "text", "text": message["content"]}
]
merged_dict = {}
for message in messages:
if message["role"] == "user":
merged_dict["role"] = message["role"]
if "content" in merged_dict:
merged_dict["content"] += message["content"]
else:
merged_dict["content"] = message["content"]
else:
processed_messages.append(message)
processed_messages.append(merged_dict)
messages = processed_messages
if self.use_default_sys_prompt:
messages = [self._generate_default_sys_prompt()] + messages
body = {
"model": self.model_id,
"messages": messages,
"temperature": self.temperature,
"max_tokens": self.max_tokens,
}
if self.response_format != "text":
body["response_format"] = {"type": self.response_format}
return body
def _generate_default_sys_prompt(self) -> Dict:
loc = self._get_location()
os = self._get_linux_distribution()
current_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
promt_str = BASIC_SYS_PROMPT.format(loc, os, current_time)
return {"role": "system", "content": promt_str}
def _get_linux_distribution(self) -> str:
platform = sysconfig.get_platform()
if "linux" in platform:
if os.path.exists("/etc/lsb-release"):
with open("/etc/lsb-release", "r") as f:
for line in f:
if line.startswith("DISTRIB_DESCRIPTION="):
return line.split("=")[1].strip()
elif os.path.exists("/etc/os-release"):
with open("/etc/os-release", "r") as f:
for line in f:
if line.startswith("PRETTY_NAME="):
return line.split("=")[1].strip()
return platform
def _get_location(self) -> str:
g = geocoder.ip("me")
if g.ok:
return g.city + "," + g.country
else:
return "unknown"
|