# ========= 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 os
import subprocess
from typing import Any, Dict, List, Optional, Type, Union
from openai import AsyncOpenAI, AsyncStream, OpenAI, Stream
from pydantic import BaseModel
from camel.configs import OLLAMA_API_PARAMS, OllamaConfig
from camel.messages import OpenAIMessage
from camel.models import BaseModelBackend
from camel.models._utils import try_modify_message_with_format
from camel.types import (
ChatCompletion,
ChatCompletionChunk,
ModelType,
)
from camel.utils import BaseTokenCounter, OpenAITokenCounter
[docs]
class OllamaModel(BaseModelBackend):
r"""Ollama service interface.
Args:
model_type (Union[ModelType, str]): Model for which a backend is
created.
model_config_dict (Optional[Dict[str, Any]], optional): A dictionary
that will be fed into:obj:`openai.ChatCompletion.create()`.
If:obj:`None`, :obj:`OllamaConfig().as_dict()` will be used.
(default: :obj:`None`)
api_key (Optional[str], optional): The API key for authenticating with
the model service. Ollama doesn't need API key, it would be
ignored if set. (default: :obj:`None`)
url (Optional[str], optional): The url to the model service.
(default: :obj:`None`)
token_counter (Optional[BaseTokenCounter], optional): Token counter to
use for the model. If not provided, :obj:`OpenAITokenCounter(
ModelType.GPT_4O_MINI)` will be used.
(default: :obj:`None`)
References:
https://github.com/ollama/ollama/blob/main/docs/openai.md
"""
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,
) -> None:
if model_config_dict is None:
model_config_dict = OllamaConfig().as_dict()
url = url or os.environ.get("OLLAMA_BASE_URL")
super().__init__(
model_type, model_config_dict, api_key, url, token_counter
)
if not self._url:
self._start_server()
# Use OpenAI client as interface call Ollama
self._client = OpenAI(
timeout=180,
max_retries=3,
api_key="Set-but-ignored", # required but ignored
base_url=self._url,
)
self._async_client = AsyncOpenAI(
timeout=180,
max_retries=3,
api_key="Set-but-ignored", # required but ignored
base_url=self._url,
)
def _start_server(self) -> None:
r"""Starts the Ollama server in a subprocess."""
try:
subprocess.Popen(
["ollama", "server", "--port", "11434"],
stdout=subprocess.PIPE,
stderr=subprocess.PIPE,
)
self._url = "http://localhost:11434/v1"
print(
f"Ollama server started on {self._url} "
f"for {self.model_type} model."
)
except Exception as e:
print(f"Failed to start Ollama server: {e}.")
@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(ModelType.GPT_4O_MINI)
return self._token_counter
[docs]
def check_model_config(self):
r"""Check whether the model configuration contains any
unexpected arguments to Ollama API.
Raises:
ValueError: If the model configuration dictionary contains any
unexpected arguments to OpenAI API.
"""
for param in self.model_config_dict:
if param not in OLLAMA_API_PARAMS:
raise ValueError(
f"Unexpected argument `{param}` is "
"input into Ollama model backend."
)
def _run(
self,
messages: List[OpenAIMessage],
response_format: Optional[Type[BaseModel]] = None,
tools: Optional[List[Dict[str, Any]]] = None,
) -> Union[ChatCompletion, Stream[ChatCompletionChunk]]:
r"""Runs inference of Ollama chat completion.
Args:
messages (List[OpenAIMessage]): Message list with the chat history
in OpenAI API format.
response_format (Optional[Type[BaseModel]]): The format of the
response.
tools (Optional[List[Dict[str, Any]]]): The schema of the tools to
use for the request.
Returns:
Union[ChatCompletion, Stream[ChatCompletionChunk]]:
`ChatCompletion` in the non-stream mode, or
`Stream[ChatCompletionChunk]` in the stream mode.
"""
response_format = response_format or self.model_config_dict.get(
"response_format", None
)
# For Ollama, the tool calling will be broken with response_format
if response_format and not tools:
return self._request_parse(messages, response_format, tools)
else:
return self._request_chat_completion(
messages, response_format, tools
)
async def _arun(
self,
messages: List[OpenAIMessage],
response_format: Optional[Type[BaseModel]] = None,
tools: Optional[List[Dict[str, Any]]] = None,
) -> Union[ChatCompletion, AsyncStream[ChatCompletionChunk]]:
r"""Runs inference of Ollama chat completion in async mode.
Args:
messages (List[OpenAIMessage]): Message list with the chat history
in OpenAI API format.
response_format (Optional[Type[BaseModel]]): The format of the
response.
tools (Optional[List[Dict[str, Any]]]): The schema of the tools to
use for the request.
Returns:
Union[ChatCompletion, AsyncStream[ChatCompletionChunk]]:
`ChatCompletion` in the non-stream mode, or
`AsyncStream[ChatCompletionChunk]` in the stream mode.
"""
response_format = response_format or self.model_config_dict.get(
"response_format", None
)
if response_format:
return await self._arequest_parse(messages, response_format, tools)
else:
return await self._arequest_chat_completion(
messages, response_format, tools
)
def _prepare_chat_completion_config(
self,
messages: List[OpenAIMessage],
response_format: Optional[Type[BaseModel]] = None,
tools: Optional[List[Dict[str, Any]]] = None,
) -> Dict[str, Any]:
request_config = self.model_config_dict.copy()
if tools:
request_config["tools"] = tools
if response_format:
try_modify_message_with_format(messages[-1], response_format)
request_config["response_format"] = {"type": "json_object"}
return request_config
def _request_chat_completion(
self,
messages: List[OpenAIMessage],
response_format: Optional[Type[BaseModel]] = None,
tools: Optional[List[Dict[str, Any]]] = None,
) -> Union[ChatCompletion, Stream[ChatCompletionChunk]]:
request_config = self._prepare_chat_completion_config(
messages, response_format, tools
)
return self._client.chat.completions.create(
messages=messages,
model=self.model_type,
**request_config,
)
async def _arequest_chat_completion(
self,
messages: List[OpenAIMessage],
response_format: Optional[Type[BaseModel]] = None,
tools: Optional[List[Dict[str, Any]]] = None,
) -> Union[ChatCompletion, AsyncStream[ChatCompletionChunk]]:
request_config = self._prepare_chat_completion_config(
messages, response_format, tools
)
return await self._async_client.chat.completions.create(
messages=messages,
model=self.model_type,
**request_config,
)
def _request_parse(
self,
messages: List[OpenAIMessage],
response_format: Type[BaseModel],
tools: Optional[List[Dict[str, Any]]] = None,
) -> ChatCompletion:
request_config = self.model_config_dict.copy()
request_config["response_format"] = response_format
request_config.pop("stream", None)
if tools is not None:
request_config["tools"] = tools
return self._client.beta.chat.completions.parse(
messages=messages,
model=self.model_type,
**request_config,
)
async def _arequest_parse(
self,
messages: List[OpenAIMessage],
response_format: Type[BaseModel],
tools: Optional[List[Dict[str, Any]]] = None,
) -> ChatCompletion:
request_config = self.model_config_dict.copy()
request_config["response_format"] = response_format
request_config.pop("stream", None)
if tools is not None:
request_config["tools"] = tools
return await self._async_client.beta.chat.completions.parse(
messages=messages,
model=self.model_type,
**request_config,
)
@property
def stream(self) -> bool:
r"""Returns whether the model is in stream mode, which sends partial
results each time.
Returns:
bool: Whether the model is in stream mode.
"""
return self.model_config_dict.get('stream', False)
