camel-ai 0.1.5.5__py3-none-any.whl → 0.1.5.9__py3-none-any.whl

camel/configs/vllm_config.py

@@ -0,0 +1,103 @@
+# =========== Copyright 2023 @ 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 @ CAMEL-AI.org. All Rights Reserved. ===========
+from __future__ import annotations
+
+from dataclasses import asdict, dataclass, field
+from typing import Sequence
+
+from openai._types import NOT_GIVEN, NotGiven
+
+from camel.configs.base_config import BaseConfig
+
+
+# flake8: noqa: E501
+@dataclass(frozen=True)
+class VLLMConfig(BaseConfig):
+    r"""Defines the parameters for generating chat completions using the
+    OpenAI API.
+
+    Reference: https://docs.vllm.ai/en/latest/serving/openai_compatible_server.html
+
+    Args:
+        temperature (float, optional): Sampling temperature to use, between
+            :obj:`0` and :obj:`2`. Higher values make the output more random,
+            while lower values make it more focused and deterministic.
+            (default: :obj:`0.2`)
+        top_p (float, optional): An alternative to sampling with temperature,
+            called nucleus sampling, where the model considers the results of
+            the tokens with top_p probability mass. So :obj:`0.1` means only
+            the tokens comprising the top 10% probability mass are considered.
+            (default: :obj:`1.0`)
+        n (int, optional): How many chat completion choices to generate for
+            each input message. (default: :obj:`1`)
+        response_format (object, optional): An object specifying the format
+            that the model must output. Compatible with GPT-4 Turbo and all
+            GPT-3.5 Turbo models newer than gpt-3.5-turbo-1106. Setting to
+            {"type": "json_object"} enables JSON mode, which guarantees the
+            message the model generates is valid JSON. Important: when using
+            JSON mode, you must also instruct the model to produce JSON
+            yourself via a system or user message. Without this, the model
+            may generate an unending stream of whitespace until the generation
+            reaches the token limit, resulting in a long-running and seemingly
+            "stuck" request. Also note that the message content may be
+            partially cut off if finish_reason="length", which indicates the
+            generation exceeded max_tokens or the conversation exceeded the
+            max context length.
+        stream (bool, optional): If True, partial message deltas will be sent
+            as data-only server-sent events as they become available.
+            (default: :obj:`False`)
+        stop (str or list, optional): Up to :obj:`4` sequences where the API
+            will stop generating further tokens. (default: :obj:`None`)
+        max_tokens (int, optional): The maximum number of tokens to generate
+            in the chat completion. The total length of input tokens and
+            generated tokens is limited by the model's context length.
+            (default: :obj:`None`)
+        presence_penalty (float, optional): Number between :obj:`-2.0` and
+            :obj:`2.0`. Positive values penalize new tokens based on whether
+            they appear in the text so far, increasing the model's likelihood
+            to talk about new topics. See more information about frequency and
+            presence penalties. (default: :obj:`0.0`)
+        frequency_penalty (float, optional): Number between :obj:`-2.0` and
+            :obj:`2.0`. Positive values penalize new tokens based on their
+            existing frequency in the text so far, decreasing the model's
+            likelihood to repeat the same line verbatim. See more information
+            about frequency and presence penalties. (default: :obj:`0.0`)
+        logit_bias (dict, optional): Modify the likelihood of specified tokens
+            appearing in the completion. Accepts a json object that maps tokens
+            (specified by their token ID in the tokenizer) to an associated
+            bias value from :obj:`-100` to :obj:`100`. Mathematically, the bias
+            is added to the logits generated by the model prior to sampling.
+            The exact effect will vary per model, but values between:obj:` -1`
+            and :obj:`1` should decrease or increase likelihood of selection;
+            values like :obj:`-100` or :obj:`100` should result in a ban or
+            exclusive selection of the relevant token. (default: :obj:`{}`)
+        user (str, optional): A unique identifier representing your end-user,
+            which can help OpenAI to monitor and detect abuse.
+            (default: :obj:`""`)
+    """
+
+    temperature: float = 0.2  # openai default: 1.0
+    top_p: float = 1.0
+    n: int = 1
+    stream: bool = False
+    stop: str | Sequence[str] | NotGiven = NOT_GIVEN
+    max_tokens: int | NotGiven = NOT_GIVEN
+    presence_penalty: float = 0.0
+    response_format: dict | NotGiven = NOT_GIVEN
+    frequency_penalty: float = 0.0
+    logit_bias: dict = field(default_factory=dict)
+    user: str = ""
+
+
+VLLM_API_PARAMS = {param for param in asdict(VLLMConfig()).keys()}

camel/configs/zhipuai_config.py

@@ -21,7 +21,7 @@ from openai._types import NOT_GIVEN, NotGiven
 from camel.configs.base_config import BaseConfig
 
 if TYPE_CHECKING:
-    from camel.functions import OpenAIFunction
+    from camel.toolkits import OpenAIFunction
 
 
 @dataclass(frozen=True)

camel/embeddings/__init__.py

@@ -12,6 +12,7 @@
 # limitations under the License.
 # =========== Copyright 2023 @ CAMEL-AI.org. All Rights Reserved. ===========
 from .base import BaseEmbedding
+from .mistral_embedding import MistralEmbedding
 from .openai_embedding import OpenAIEmbedding
 from .sentence_transformers_embeddings import SentenceTransformerEncoder
 from .vlm_embedding import VisionLanguageEmbedding
@@ -21,4 +22,5 @@ __all__ = [
     "OpenAIEmbedding",
     "SentenceTransformerEncoder",
     "VisionLanguageEmbedding",
+    "MistralEmbedding",
 ]

camel/embeddings/mistral_embedding.py

@@ -0,0 +1,89 @@
+# =========== Copyright 2023 @ 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 @ CAMEL-AI.org. All Rights Reserved. ===========
+from __future__ import annotations
+
+import os
+from typing import Any
+
+from camel.embeddings.base import BaseEmbedding
+from camel.types import EmbeddingModelType
+from camel.utils import api_keys_required
+
+
+class MistralEmbedding(BaseEmbedding[str]):
+    r"""Provides text embedding functionalities using Mistral's models.
+
+    Args:
+        model_type (EmbeddingModelType, optional): The model type to be
+            used for text embeddings.
+            (default: :obj:`MISTRAL_EMBED`)
+        api_key (str, optional): The API key for authenticating with the
+            Mistral service. (default: :obj:`None`)
+        dimensions (int, optional): The text embedding output dimensions.
+            (default: :obj:`None`)
+
+    Raises:
+        RuntimeError: If an unsupported model type is specified.
+    """
+
+    def __init__(
+        self,
+        model_type: EmbeddingModelType = (EmbeddingModelType.MISTRAL_EMBED),
+        api_key: str | None = None,
+        dimensions: int | None = None,
+    ) -> None:
+        from mistralai.client import MistralClient
+
+        if not model_type.is_mistral:
+            raise ValueError("Invalid Mistral embedding model type.")
+        self.model_type = model_type
+        if dimensions is None:
+            self.output_dim = model_type.output_dim
+        else:
+            assert isinstance(dimensions, int)
+            self.output_dim = dimensions
+        self._api_key = api_key or os.environ.get("MISTRAL_API_KEY")
+        self._client = MistralClient(api_key=self._api_key)
+
+    @api_keys_required("MISTRAL_API_KEY")
+    def embed_list(
+        self,
+        objs: list[str],
+        **kwargs: Any,
+    ) -> list[list[float]]:
+        r"""Generates embeddings for the given texts.
+
+        Args:
+            objs (list[str]): The texts for which to generate the embeddings.
+            **kwargs (Any): Extra kwargs passed to the embedding API.
+
+        Returns:
+            list[list[float]]: A list that represents the generated embedding
+                as a list of floating-point numbers.
+        """
+        # TODO: count tokens
+        response = self._client.embeddings(
+            input=objs,
+            model=self.model_type.value,
+            **kwargs,
+        )
+        return [data.embedding for data in response.data]
+
+    def get_output_dim(self) -> int:
+        r"""Returns the output dimension of the embeddings.
+
+        Returns:
+            int: The dimensionality of the embedding for the current model.
+        """
+        return self.output_dim

camel/interpreters/__init__.py

@@ -16,6 +16,7 @@ from .base import BaseInterpreter
 from .docker_interpreter import DockerInterpreter
 from .internal_python_interpreter import InternalPythonInterpreter
 from .interpreter_error import InterpreterError
+from .ipython_interpreter import JupyterKernelInterpreter
 from .subprocess_interpreter import SubprocessInterpreter
 
 __all__ = [
@@ -24,4 +25,5 @@ __all__ = [
     'InternalPythonInterpreter',
     'SubprocessInterpreter',
     'DockerInterpreter',
+    'JupyterKernelInterpreter',
 ]

camel/interpreters/ipython_interpreter.py

@@ -0,0 +1,167 @@
+# =========== Copyright 2023 @ 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 @ CAMEL-AI.org. All Rights Reserved. ===========
+
+import queue
+import re
+from typing import TYPE_CHECKING, Any, Dict, List, Optional
+
+from camel.interpreters.base import BaseInterpreter
+from camel.interpreters.interpreter_error import InterpreterError
+
+if TYPE_CHECKING:
+    from jupyter_client import BlockingKernelClient, KernelManager
+
+TIMEOUT = 30
+
+
+class JupyterKernelInterpreter(BaseInterpreter):
+    r"""A class for executing code strings in a Jupyter Kernel.
+
+    Args:
+        require_confirm (bool, optional): If `True`, prompt user before
+            running code strings for security. Defaults to `True`.
+        print_stdout (bool, optional): If `True`, print the standard
+            output of the executed code. Defaults to `False`.
+        print_stderr (bool, optional): If `True`, print the standard error
+            of the executed code. Defaults to `True`.
+    """
+
+    def __init__(
+        self,
+        require_confirm: bool = True,
+        print_stdout: bool = False,
+        print_stderr: bool = True,
+    ) -> None:
+        self.require_confirm = require_confirm
+        self.print_stdout = print_stdout
+        self.print_stderr = print_stderr
+
+        self.kernel_manager: Optional[KernelManager] = None
+        self.client: Optional[BlockingKernelClient] = None
+
+    def __del__(self) -> None:
+        r"""Clean up the kernel and client."""
+
+        if self.kernel_manager:
+            self.kernel_manager.shutdown_kernel()
+        if self.client:
+            self.client.stop_channels()
+
+    def _initialize_if_needed(self) -> None:
+        r"""Initialize the kernel manager and client if they are not already
+        initialized.
+        """
+
+        if self.kernel_manager is not None:
+            return
+
+        from jupyter_client.manager import start_new_kernel
+
+        self.kernel_manager, self.client = start_new_kernel()
+
+    @staticmethod
+    def _clean_ipython_output(output: str) -> str:
+        r"""Remove ANSI escape sequences from the output."""
+
+        ansi_escape = re.compile(r'\x1B[@-_][0-?]*[ -/]*[@-~]')
+        return ansi_escape.sub('', output)
+
+    def _execute(self, code: str, timeout: float) -> str:
+        r"""Execute the code in the Jupyter kernel and return the result."""
+
+        if not self.kernel_manager or not self.client:
+            raise InterpreterError("Jupyter client is not initialized.")
+
+        self.client.execute(code)
+        outputs = []
+        while True:
+            try:
+                msg = self.client.get_iopub_msg(timeout=timeout)
+                msg_content = msg["content"]
+                msg_type = msg.get("msg_type", None)
+
+                if msg_content.get("execution_state", None) == "idle":
+                    break
+
+                if msg_type == "error":
+                    print(msg_content.keys())
+                    print(msg_content)
+                    traceback = "\n".join(msg_content["traceback"])
+                    outputs.append(traceback)
+                elif msg_type == "stream":
+                    outputs.append(msg_content["text"])
+                elif msg_type in ["execute_result", "display_data"]:
+                    outputs.append(msg_content["data"]["text/plain"])
+                    if "image/png" in msg_content["data"]:
+                        outputs.append(
+                            f"\n![image](data:image/png;base64,{msg_content['data']['image/png']})\n"
+                        )
+            except queue.Empty:
+                outputs.append("Time out")
+                break
+            except Exception as e:
+                outputs.append(f"Exception occurred: {e!s}")
+                break
+
+        exec_result = "\n".join(outputs)
+        return self._clean_ipython_output(exec_result)
+
+    def run(self, code: str, code_type: str) -> str:
+        r"""Executes the given code in the Jupyter kernel.
+
+        Args:
+            code (str): The code string to execute.
+            code_type (str): The type of code to execute (e.g., 'python',
+                'bash').
+
+        Returns:
+            str: A string containing the captured result of the
+                executed code.
+
+        Raises:
+            InterpreterError: If there is an error when doing code execution.
+        """
+        self._initialize_if_needed()
+
+        if code_type == "bash":
+            code = f"%%bash\n({code})"
+        try:
+            result = self._execute(code, timeout=TIMEOUT)
+        except Exception as e:
+            raise InterpreterError(f"Execution failed: {e!s}")
+
+        return result
+
+    def supported_code_types(self) -> List[str]:
+        r"""Provides supported code types by the interpreter.
+
+        Returns:
+            List[str]: Supported code types.
+        """
+        return ["python", "bash"]
+
+    def update_action_space(self, action_space: Dict[str, Any]) -> None:
+        r"""Updates the action space for the interpreter.
+
+        Args:
+            action_space (Dict[str, Any]): A dictionary representing the
+                new or updated action space.
+
+        Raises:
+            RuntimeError: Always raised because `JupyterKernelInterpreter`
+                does not support updating the action space.
+        """
+        raise RuntimeError(
+            "SubprocessInterpreter doesn't support " "`action_space`."
+        )

camel/models/__init__.py

@@ -12,8 +12,12 @@
 # limitations under the License.
 # =========== Copyright 2023 @ CAMEL-AI.org. All Rights Reserved. ===========
 from .anthropic_model import AnthropicModel
+from .azure_openai_model import AzureOpenAIModel
 from .base_model import BaseModelBackend
+from .gemini_model import GeminiModel
+from .groq_model import GroqModel
 from .litellm_model import LiteLLMModel
+from .mistral_model import MistralModel
 from .model_factory import ModelFactory
 from .nemotron_model import NemotronModel
 from .ollama_model import OllamaModel
@@ -21,12 +25,16 @@ from .open_source_model import OpenSourceModel
 from .openai_audio_models import OpenAIAudioModels
 from .openai_model import OpenAIModel
 from .stub_model import StubModel
+from .vllm_model import VLLMModel
 from .zhipuai_model import ZhipuAIModel
 
 __all__ = [
     'BaseModelBackend',
     'OpenAIModel',
+    'AzureOpenAIModel',
     'AnthropicModel',
+    'MistralModel',
+    'GroqModel',
     'StubModel',
     'ZhipuAIModel',
     'OpenSourceModel',
@@ -35,4 +43,6 @@ __all__ = [
     'OpenAIAudioModels',
     'NemotronModel',
     'OllamaModel',
+    'VLLMModel',
+    'GeminiModel',
 ]

camel/models/anthropic_model.py

@@ -36,6 +36,7 @@ class AnthropicModel(BaseModelBackend):
         model_config_dict: Dict[str, Any],
         api_key: Optional[str] = None,
         url: Optional[str] = None,
+        token_counter: Optional[BaseTokenCounter] = None,
     ) -> None:
         r"""Constructor for Anthropic backend.
 
@@ -48,12 +49,16 @@ class AnthropicModel(BaseModelBackend):
                 Anthropic service. (default: :obj:`None`)
             url (Optional[str]): The url to the Anthropic service. (default:
                 :obj:`None`)
+            token_counter (Optional[BaseTokenCounter]): Token counter to use
+                for the model. If not provided, `AnthropicTokenCounter` will
+                be used.
         """
-        super().__init__(model_type, model_config_dict, api_key, url)
+        super().__init__(
+            model_type, model_config_dict, api_key, url, token_counter
+        )
         self._api_key = api_key or os.environ.get("ANTHROPIC_API_KEY")
         self._url = url or os.environ.get("ANTHROPIC_API_BASE_URL")
         self.client = Anthropic(api_key=self._api_key, base_url=self._url)
-        self._token_counter: Optional[BaseTokenCounter] = None
 
     def _convert_response_from_anthropic_to_openai(self, response):
         # openai ^1.0.0 format, reference openai/types/chat/chat_completion.py

camel/models/azure_openai_model.py

@@ -0,0 +1,152 @@
+# =========== Copyright 2023 @ 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 @ CAMEL-AI.org. All Rights Reserved. ===========
+import os
+from typing import Any, Dict, List, Optional, Union
+
+from openai import AzureOpenAI, Stream
+
+from camel.configs import OPENAI_API_PARAMS
+from camel.messages import OpenAIMessage
+from camel.models.base_model import BaseModelBackend
+from camel.types import ChatCompletion, ChatCompletionChunk, ModelType
+from camel.utils import BaseTokenCounter, OpenAITokenCounter, api_keys_required
+
+
+class AzureOpenAIModel(BaseModelBackend):
+    r"""Azure OpenAI API in a unified BaseModelBackend interface.
+    Doc: https://learn.microsoft.com/en-us/azure/ai-services/openai/
+    """
+
+    def __init__(
+        self,
+        model_type: ModelType,
+        model_config_dict: Dict[str, Any],
+        api_key: Optional[str] = None,
+        url: Optional[str] = None,
+        api_version: Optional[str] = None,
+        azure_deployment_name: Optional[str] = None,
+    ) -> None:
+        r"""Constructor for OpenAI backend.
+
+        Args:
+            model_type (ModelType): Model for which a backend is created,
+                one of GPT_* series.
+            model_config_dict (Dict[str, Any]): A dictionary that will
+                be fed into openai.ChatCompletion.create().
+            api_key (Optional[str]): The API key for authenticating with the
+                OpenAI service. (default: :obj:`None`)
+            url (Optional[str]): The url to the OpenAI service. (default:
+                :obj:`None`)
+            api_version (Optional[str]): The api version for the model.
+            azure_deployment_name (Optional[str]): The deployment name you
+                chose when you deployed an azure model. (default: :obj:`None`)
+        """
+        super().__init__(model_type, model_config_dict, api_key, url)
+        self._url = url or os.environ.get("AZURE_OPENAI_ENDPOINT")
+        self._api_key = api_key or os.environ.get("AZURE_OPENAI_API_KEY")
+        self.api_version = api_version or os.environ.get("AZURE_API_VERSION")
+        self.azure_deployment_name = azure_deployment_name or os.environ.get(
+            "AZURE_DEPLOYMENT_NAME"
+        )
+
+        if self._url is None:
+            raise ValueError(
+                "Must provide either the `url` argument "
+                "or `AZURE_OPENAI_ENDPOINT` environment variable."
+            )
+        if self._api_key is None:
+            raise ValueError(
+                "Must provide either the `api_key` argument "
+                "or `AZURE_OPENAI_API_KEY` environment variable."
+            )
+        if self.api_version is None:
+            raise ValueError(
+                "Must provide either the `api_version` argument "
+                "or `AZURE_API_VERSION` environment variable."
+            )
+        if self.azure_deployment_name is None:
+            raise ValueError(
+                "Must provide either the `azure_deployment_name` argument "
+                "or `AZURE_DEPLOYMENT_NAME` environment variable."
+            )
+        self.model = str(self.azure_deployment_name)
+
+        self._client = AzureOpenAI(
+            azure_endpoint=str(self._url),
+            azure_deployment=self.azure_deployment_name,
+            api_version=self.api_version,
+            api_key=self._api_key,
+            timeout=60,
+            max_retries=3,
+        )
+        self._token_counter: Optional[BaseTokenCounter] = None
+
+    @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(self.model_type)
+        return self._token_counter
+
+    @api_keys_required("AZURE_OPENAI_API_KEY", "AZURE_API_VERSION")
+    def run(
+        self,
+        messages: List[OpenAIMessage],
+    ) -> Union[ChatCompletion, Stream[ChatCompletionChunk]]:
+        r"""Runs inference of Azure OpenAI chat completion.
+
+        Args:
+            messages (List[OpenAIMessage]): Message list with the chat history
+                in OpenAI API format.
+
+        Returns:
+            Union[ChatCompletion, Stream[ChatCompletionChunk]]:
+                `ChatCompletion` in the non-stream mode, or
+                `Stream[ChatCompletionChunk]` in the stream mode.
+        """
+        response = self._client.chat.completions.create(
+            messages=messages,
+            model=self.model,
+            **self.model_config_dict,
+        )
+        return response
+
+    def check_model_config(self):
+        r"""Check whether the model configuration contains any
+        unexpected arguments to Azure OpenAI API.
+
+        Raises:
+            ValueError: If the model configuration dictionary contains any
+                unexpected arguments to Azure OpenAI API.
+        """
+        for param in self.model_config_dict:
+            if param not in OPENAI_API_PARAMS:
+                raise ValueError(
+                    f"Unexpected argument `{param}` is "
+                    "input into Azure OpenAI model backend."
+                )
+
+    @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)

camel/models/base_model.py

@@ -32,6 +32,7 @@ class BaseModelBackend(ABC):
         model_config_dict: Dict[str, Any],
         api_key: Optional[str] = None,
         url: Optional[str] = None,
+        token_counter: Optional[BaseTokenCounter] = None,
     ) -> None:
         r"""Constructor for the model backend.
 
@@ -41,13 +42,16 @@ class BaseModelBackend(ABC):
             api_key (Optional[str]): The API key for authenticating with the
                 model service.
             url (Optional[str]): The url to the model service.
+            token_counter (Optional[BaseTokenCounter]): Token counter to use
+                for the model. If not provided, `OpenAITokenCounter` will
+                be used.
         """
         self.model_type = model_type
-
         self.model_config_dict = model_config_dict
         self._api_key = api_key
         self._url = url
         self.check_model_config()
+        self._token_counter = token_counter
 
     @property
     @abstractmethod
@@ -109,7 +113,10 @@ class BaseModelBackend(ABC):
         Returns:
             int: The maximum token limit for the given model.
         """
-        return self.model_type.token_limit
+        return (
+            self.model_config_dict.get("max_tokens")
+            or self.model_type.token_limit
+        )
 
     @property
     def stream(self) -> bool: