File size: 10,320 Bytes
62da328 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 |
# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
import ast
import json
import logging
import os
import uuid
from typing import TYPE_CHECKING, Any, Dict, List, Optional, Union
if TYPE_CHECKING:
from cohere.types import ChatMessageV2, ChatResponse
from camel.configs import COHERE_API_PARAMS, CohereConfig
from camel.messages import OpenAIMessage
from camel.models import BaseModelBackend
from camel.types import ChatCompletion, ModelType
from camel.utils import (
BaseTokenCounter,
OpenAITokenCounter,
api_keys_required,
)
try:
if os.getenv("AGENTOPS_API_KEY") is not None:
from agentops import LLMEvent, record
else:
raise ImportError
except (ImportError, AttributeError):
LLMEvent = None
class CohereModel(BaseModelBackend):
r"""Cohere API in a unified BaseModelBackend interface."""
def __init__(
self,
model_type: Union[ModelType, str],
model_config_dict: Optional[Dict[str, Any]] = None,
api_key: Optional[str] = None,
url: Optional[str] = None,
token_counter: Optional[BaseTokenCounter] = None,
):
import cohere
if model_config_dict is None:
model_config_dict = CohereConfig().as_dict()
api_key = api_key or os.environ.get("COHERE_API_KEY")
url = url or os.environ.get("COHERE_API_BASE_URL")
super().__init__(
model_type, model_config_dict, api_key, url, token_counter
)
self._client = cohere.ClientV2(api_key=self._api_key)
def _to_openai_response(self, response: 'ChatResponse') -> ChatCompletion:
if response.usage and response.usage.tokens:
input_tokens = response.usage.tokens.input_tokens or 0
output_tokens = response.usage.tokens.output_tokens or 0
usage = {
"prompt_tokens": input_tokens,
"completion_tokens": output_tokens,
"total_tokens": input_tokens + output_tokens,
}
else:
usage = {}
tool_calls = response.message.tool_calls
choices = []
if tool_calls:
for tool_call in tool_calls:
openai_tool_calls = [
dict(
id=tool_call.id,
function={
"name": tool_call.function.name,
"arguments": tool_call.function.arguments,
}
if tool_call.function
else {},
type=tool_call.type,
)
]
choice = dict(
index=None,
message={
"role": "assistant",
"content": response.message.tool_plan,
"tool_calls": openai_tool_calls,
},
finish_reason=response.finish_reason
if response.finish_reason
else None,
)
choices.append(choice)
else:
openai_tool_calls = None
choice = dict(
index=None,
message={
"role": "assistant",
"content": response.message.content[0].text, # type: ignore[union-attr,index]
"tool_calls": openai_tool_calls,
},
finish_reason=response.finish_reason
if response.finish_reason
else None,
)
choices.append(choice)
obj = ChatCompletion.construct(
id=response.id,
choices=choices,
created=None,
model=self.model_type,
object="chat.completion",
usage=usage,
)
return obj
def _to_cohere_chatmessage(
self, messages: List[OpenAIMessage]
) -> List["ChatMessageV2"]:
from cohere.types import ToolCallV2Function
from cohere.types.chat_message_v2 import (
AssistantChatMessageV2,
SystemChatMessageV2,
ToolCallV2,
ToolChatMessageV2,
UserChatMessageV2,
)
tool_call_id = None
new_messages = []
for msg in messages:
role = msg.get("role")
content = msg.get("content")
function_call = msg.get("function_call")
if role == "user":
new_message = UserChatMessageV2(role="user", content=content) # type: ignore[arg-type]
elif role in {"tool", "function"}:
new_message = ToolChatMessageV2(
role="tool",
tool_call_id=tool_call_id, # type: ignore[arg-type]
content=content, # type: ignore[assignment,arg-type]
)
elif role == "assistant":
if not function_call:
new_message = AssistantChatMessageV2( # type: ignore[assignment]
role="assistant",
content=content, # type: ignore[arg-type]
)
else:
arguments = function_call.get("arguments") # type: ignore[attr-defined]
arguments_dict = ast.literal_eval(arguments)
arguments_json = json.dumps(arguments_dict)
assis_tool_call_id = str(uuid.uuid4())
tool_call_id = assis_tool_call_id
new_message = AssistantChatMessageV2( # type: ignore[assignment]
role="assistant",
tool_calls=[
ToolCallV2(
id=assis_tool_call_id,
type="function",
function=ToolCallV2Function(
name=function_call.get("name"), # type: ignore[attr-defined]
arguments=arguments_json, # type: ignore[attr-defined]
),
)
],
content=content, # type: ignore[arg-type]
)
elif role == "system":
new_message = SystemChatMessageV2( # type: ignore[assignment]
role="system",
content=content, # type: ignore[arg-type]
)
else:
raise ValueError(f"Unsupported message role: {role}")
new_messages.append(new_message)
return new_messages # type: ignore[return-value]
@property
def token_counter(self) -> BaseTokenCounter:
r"""Initialize the token counter for the model backend.
Returns:
BaseTokenCounter: The token counter following the model's
tokenization style.
"""
if not self._token_counter:
self._token_counter = OpenAITokenCounter(
model=ModelType.GPT_4O_MINI
)
return self._token_counter
@api_keys_required("COHERE_API_KEY")
def run(self, messages: List[OpenAIMessage]) -> ChatCompletion:
r"""Runs inference of Cohere chat completion.
Args:
messages (List[OpenAIMessage]): Message list with the chat history
in OpenAI API format.
Returns:
ChatCompletion.
"""
from cohere.core.api_error import ApiError
cohere_messages = self._to_cohere_chatmessage(messages)
try:
response = self._client.chat(
messages=cohere_messages,
model=self.model_type,
**self.model_config_dict,
)
except ApiError as e:
logging.error(f"Cohere API Error: {e.status_code}")
logging.error(f"Error body: {e.body}")
raise
except Exception as e:
logging.error(f"Unexpected error when calling Cohere API: {e!s}")
raise
openai_response = self._to_openai_response(response)
# Add AgentOps LLM Event tracking
if LLMEvent:
llm_event = LLMEvent(
thread_id=openai_response.id,
prompt=" ".join(
[message.get("content") for message in messages] # type: ignore[misc]
),
prompt_tokens=openai_response.usage.prompt_tokens, # type: ignore[union-attr]
completion=openai_response.choices[0].message.content,
completion_tokens=openai_response.usage.completion_tokens, # type: ignore[union-attr]
model=self.model_type,
)
record(llm_event)
return openai_response
def check_model_config(self):
r"""Check whether the model configuration contains any unexpected
arguments to Cohere API.
Raises:
ValueError: If the model configuration dictionary contains any
unexpected arguments to Cohere API.
"""
for param in self.model_config_dict:
if param not in COHERE_API_PARAMS:
raise ValueError(
f"Unexpected argument `{param}` is "
"input into Cohere model backend."
)
@property
def stream(self) -> bool:
r"""Returns whether the model is in stream mode, which sends partial
results each time. Current it's not supported.
Returns:
bool: Whether the model is in stream mode.
"""
return False
|