camel-ai 0.2.65__py3-none-any.whl → 0.2.82__py3-none-any.whl

camel/models/model_manager.py

@@ -1,4 +1,4 @@
-# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
+# ========= Copyright 2023-2025 @ 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
@@ -10,7 +10,7 @@
 # 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. =========
+# ========= Copyright 2023-2025 @ CAMEL-AI.org. All Rights Reserved. =========
 
 import asyncio
 import logging
@@ -27,6 +27,10 @@ from typing import (
 )
 
 from openai import AsyncStream, Stream
+from openai.lib.streaming.chat import (
+    AsyncChatCompletionStreamManager,
+    ChatCompletionStreamManager,
+)
 from pydantic import BaseModel
 
 from camel.messages import OpenAIMessage
@@ -196,7 +200,11 @@ class ModelManager:
         messages: List[OpenAIMessage],
         response_format: Optional[Type[BaseModel]] = None,
         tools: Optional[List[Dict[str, Any]]] = None,
-    ) -> Union[ChatCompletion, Stream[ChatCompletionChunk]]:
+    ) -> Union[
+        ChatCompletion,
+        Stream[ChatCompletionChunk],
+        ChatCompletionStreamManager[BaseModel],
+    ]:
         r"""Process a list of messages by selecting a model based on
             the scheduling strategy.
             Sends the entire list of messages to the selected model,
@@ -207,9 +215,12 @@ class ModelManager:
                 history in OpenAI API format.
 
         Returns:
-            Union[ChatCompletion, Stream[ChatCompletionChunk]]:
+            Union[ChatCompletion, Stream[ChatCompletionChunk],
+                  ChatCompletionStreamManager[BaseModel]]:
                 `ChatCompletion` in the non-stream mode, or
-                `Stream[ChatCompletionChunk]` in the stream mode.
+                `Stream[ChatCompletionChunk]` in the stream mode, or
+                `ChatCompletionStreamManager[BaseModel]` for
+                structured-output stream.
         """
         self.current_model = self.scheduling_strategy()
 
@@ -233,7 +244,11 @@ class ModelManager:
         messages: List[OpenAIMessage],
         response_format: Optional[Type[BaseModel]] = None,
         tools: Optional[List[Dict[str, Any]]] = None,
-    ) -> Union[ChatCompletion, AsyncStream[ChatCompletionChunk]]:
+    ) -> Union[
+        ChatCompletion,
+        AsyncStream[ChatCompletionChunk],
+        AsyncChatCompletionStreamManager[BaseModel],
+    ]:
         r"""Process a list of messages by selecting a model based on
             the scheduling strategy.
             Sends the entire list of messages to the selected model,
@@ -244,9 +259,12 @@ class ModelManager:
                 history in OpenAI API format.
 
         Returns:
-            Union[ChatCompletion, AsyncStream[ChatCompletionChunk]]:
+            Union[ChatCompletion, AsyncStream[ChatCompletionChunk],
+                  AsyncChatCompletionStreamManager[BaseModel]]:
                 `ChatCompletion` in the non-stream mode, or
-                `AsyncStream[ChatCompletionChunk]` in the stream mode.
+                `AsyncStream[ChatCompletionChunk]` in the stream mode, or
+                `AsyncChatCompletionStreamManager[BaseModel]` for
+                structured-output stream.
         """
         async with self.lock:
             self.current_model = self.scheduling_strategy()

camel/models/modelscope_model.py

@@ -1,4 +1,4 @@
-# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
+# ========= Copyright 2023-2025 @ 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
@@ -10,22 +10,14 @@
 # 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. =========
+# ========= Copyright 2023-2025 @ CAMEL-AI.org. All Rights Reserved. =========
 
 import os
-import time
-from typing import Any, Dict, List, Optional, Union
+from typing import Any, Dict, Optional, Union
 
-from openai import AsyncStream, Stream
-
-from camel.configs import MODELSCOPE_API_PARAMS, ModelScopeConfig
-from camel.messages import OpenAIMessage
+from camel.configs import ModelScopeConfig
 from camel.models.openai_compatible_model import OpenAICompatibleModel
-from camel.types import (
-    ChatCompletion,
-    ChatCompletionChunk,
-    ModelType,
-)
+from camel.types import ModelType
 from camel.utils import (
     BaseTokenCounter,
     api_keys_required,
@@ -56,6 +48,10 @@ class ModelScopeModel(OpenAICompatibleModel):
             API calls. If not provided, will fall back to the MODEL_TIMEOUT
             environment variable or default to 180 seconds.
             (default: :obj:`None`)
+        max_retries (int, optional): Maximum number of retries for API calls.
+            (default: :obj:`3`)
+        **kwargs (Any): Additional arguments to pass to the client
+            initialization.
     """
 
     @api_keys_required(
@@ -71,6 +67,8 @@ class ModelScopeModel(OpenAICompatibleModel):
         url: Optional[str] = None,
         token_counter: Optional[BaseTokenCounter] = None,
         timeout: Optional[float] = None,
+        max_retries: int = 3,
+        **kwargs: Any,
     ) -> None:
         if model_config_dict is None:
             model_config_dict = ModelScopeConfig().as_dict()
@@ -87,184 +85,6 @@ class ModelScopeModel(OpenAICompatibleModel):
             url=url,
             token_counter=token_counter,
             timeout=timeout,
+            max_retries=max_retries,
+            **kwargs,
         )
-
-    def _post_handle_response(
-        self, response: Union[ChatCompletion, Stream[ChatCompletionChunk]]
-    ) -> ChatCompletion:
-        r"""Handle reasoning content with <think> tags at the beginning."""
-        if not isinstance(response, Stream):
-            # Handle non-streaming response (existing logic)
-            if self.model_config_dict.get("extra_body", {}).get(
-                "enable_thinking", False
-            ):
-                reasoning_content = response.choices[
-                    0
-                ].message.reasoning_content  # type: ignore[attr-defined]
-                combined_content = (
-                    f"<think>\n{reasoning_content}\n</think>\n"
-                    if reasoning_content
-                    else ""
-                )
-                response_content = response.choices[0].message.content or ""
-                combined_content += response_content
-
-                # Construct a new ChatCompletion with combined content
-                return ChatCompletion.construct(
-                    id=response.id,
-                    choices=[
-                        dict(
-                            finish_reason=response.choices[0].finish_reason,
-                            index=response.choices[0].index,
-                            logprobs=response.choices[0].logprobs,
-                            message=dict(
-                                role=response.choices[0].message.role,
-                                content=combined_content,
-                            ),
-                        )
-                    ],
-                    created=response.created,
-                    model=response.model,
-                    object="chat.completion",
-                    system_fingerprint=response.system_fingerprint,
-                    usage=response.usage,
-                )
-            else:
-                return response  # Return original if no thinking enabled
-
-        # Handle streaming response
-        accumulated_reasoning = ""
-        accumulated_content = ""
-        final_chunk = None
-        usage_data = None  # Initialize usage data
-        role = "assistant"  # Default role
-
-        for chunk in response:
-            final_chunk = chunk  # Keep track of the last chunk for metadata
-            if chunk.choices:
-                delta = chunk.choices[0].delta
-                if delta.role:
-                    role = delta.role  # Update role if provided
-                if (
-                    hasattr(delta, 'reasoning_content')
-                    and delta.reasoning_content
-                ):
-                    accumulated_reasoning += delta.reasoning_content
-                if delta.content:
-                    accumulated_content += delta.content
-
-            if hasattr(chunk, 'usage') and chunk.usage:
-                usage_data = chunk.usage
-
-        combined_content = (
-            f"<think>\n{accumulated_reasoning}\n</think>\n"
-            if accumulated_reasoning
-            else ""
-        ) + accumulated_content
-
-        # Construct the final ChatCompletion object from accumulated
-        # stream data
-        if final_chunk:
-            finish_reason = "stop"  # Default finish reason
-            logprobs = None
-            if final_chunk.choices:
-                finish_reason = (
-                    final_chunk.choices[0].finish_reason or finish_reason
-                )
-                if hasattr(final_chunk.choices[0], 'logprobs'):
-                    logprobs = final_chunk.choices[0].logprobs
-
-            return ChatCompletion.construct(
-                # Use data from the final chunk or defaults
-                id=final_chunk.id
-                if hasattr(final_chunk, 'id')
-                else "streamed-completion",
-                choices=[
-                    dict(
-                        finish_reason=finish_reason,
-                        index=0,
-                        logprobs=logprobs,
-                        message=dict(
-                            role=role,
-                            content=combined_content,
-                        ),
-                    )
-                ],
-                created=final_chunk.created
-                if hasattr(final_chunk, 'created')
-                else int(time.time()),
-                model=final_chunk.model
-                if hasattr(final_chunk, 'model')
-                else self.model_type,
-                object="chat.completion",
-                system_fingerprint=final_chunk.system_fingerprint
-                if hasattr(final_chunk, 'system_fingerprint')
-                else None,
-                usage=usage_data,
-            )
-        else:
-            # Handle cases where the stream was empty or invalid
-            return ChatCompletion.construct(
-                id="empty-stream",
-                choices=[
-                    dict(
-                        finish_reason="error",
-                        index=0,
-                        message=dict(role="assistant", content=""),
-                    )
-                ],
-                created=int(time.time()),
-                model=self.model_type,
-                object="chat.completion",
-                usage=usage_data,
-            )
-
-    def _request_chat_completion(
-        self,
-        messages: List[OpenAIMessage],
-        tools: Optional[List[Dict[str, Any]]] = None,
-    ) -> Union[ChatCompletion, Stream[ChatCompletionChunk]]:
-        request_config = self.model_config_dict.copy()
-
-        if tools:
-            request_config["tools"] = tools
-
-        return self._post_handle_response(
-            self._client.chat.completions.create(
-                messages=messages,
-                model=self.model_type,
-                **request_config,
-            )
-        )
-
-    async def _arequest_chat_completion(
-        self,
-        messages: List[OpenAIMessage],
-        tools: Optional[List[Dict[str, Any]]] = None,
-    ) -> Union[ChatCompletion, AsyncStream[ChatCompletionChunk]]:
-        request_config = self.model_config_dict.copy()
-
-        if tools:
-            request_config["tools"] = tools
-
-        response = await self._async_client.chat.completions.create(
-            messages=messages,
-            model=self.model_type,
-            **request_config,
-        )
-        return self._post_handle_response(response)
-
-    def check_model_config(self):
-        r"""Check whether the model configuration contains any
-        unexpected arguments to ModelScope API.
-
-        Raises:
-            ValueError: If the model configuration dictionary contains any
-                unexpected arguments to ModelScope API.
-        """
-        for param in self.model_config_dict:
-            if param not in MODELSCOPE_API_PARAMS:
-                raise ValueError(
-                    f"Unexpected argument `{param}` is "
-                    "input into ModelScope model backend."
-                )

camel/models/moonshot_model.py

@@ -1,4 +1,4 @@
-# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
+# ========= Copyright 2023-2025 @ 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
@@ -10,16 +10,19 @@
 # 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. =========
+# ========= Copyright 2023-2025 @ CAMEL-AI.org. All Rights Reserved. =========
 
+import copy
 import os
 from typing import Any, Dict, List, Optional, Type, Union
 
 from openai import AsyncStream
 from pydantic import BaseModel
 
-from camel.configs import MOONSHOT_API_PARAMS, MoonshotConfig
+from camel.configs import MoonshotConfig
+from camel.logger import get_logger
 from camel.messages import OpenAIMessage
+from camel.models._utils import try_modify_message_with_format
 from camel.models.openai_compatible_model import OpenAICompatibleModel
 from camel.types import (
     ChatCompletion,
@@ -29,8 +32,25 @@ from camel.types import (
 from camel.utils import (
     BaseTokenCounter,
     api_keys_required,
+    get_current_agent_session_id,
+    update_langfuse_trace,
 )
 
+logger = get_logger(__name__)
+
+if os.environ.get("LANGFUSE_ENABLED", "False").lower() == "true":
+    try:
+        from langfuse.decorators import observe
+    except ImportError:
+        from camel.utils import observe
+elif os.environ.get("TRACEROOT_ENABLED", "False").lower() == "true":
+    try:
+        from traceroot import trace as observe  # type: ignore[import]
+    except ImportError:
+        from camel.utils import observe
+else:
+    from camel.utils import observe
+
 
 class MoonshotModel(OpenAICompatibleModel):
     r"""Moonshot API in a unified OpenAICompatibleModel interface.
@@ -45,7 +65,9 @@ class MoonshotModel(OpenAICompatibleModel):
         api_key (Optional[str], optional): The API key for authenticating with
             the Moonshot service. (default: :obj:`None`)
         url (Optional[str], optional): The url to the Moonshot service.
-            (default: :obj:`https://api.moonshot.cn/v1`)
+            For Chinese users, use :obj:`https://api.moonshot.cn/v1`.
+            For overseas users, the default endpoint will be used.
+            (default: :obj:`https://api.moonshot.ai/v1`)
         token_counter (Optional[BaseTokenCounter], optional): Token counter to
             use for the model. If not provided, :obj:`OpenAITokenCounter(
             ModelType.GPT_4)` will be used.
@@ -54,6 +76,10 @@ class MoonshotModel(OpenAICompatibleModel):
             API calls. If not provided, will fall back to the MODEL_TIMEOUT
             environment variable or default to 180 seconds.
             (default: :obj:`None`)
+        max_retries (int, optional): Maximum number of retries for API calls.
+            (default: :obj:`3`)
+        **kwargs (Any): Additional arguments to pass to the client
+            initialization.
     """
 
     @api_keys_required([("api_key", "MOONSHOT_API_KEY")])
@@ -65,14 +91,18 @@ class MoonshotModel(OpenAICompatibleModel):
         url: Optional[str] = None,
         token_counter: Optional[BaseTokenCounter] = None,
         timeout: Optional[float] = None,
+        max_retries: int = 3,
+        **kwargs: Any,
     ) -> None:
         if model_config_dict is None:
             model_config_dict = MoonshotConfig().as_dict()
         api_key = api_key or os.environ.get("MOONSHOT_API_KEY")
-        url = url or os.environ.get(
-            "MOONSHOT_API_BASE_URL",
-            "https://api.moonshot.cn/v1",
-        )
+        # Preserve default URL if not provided
+        if url is None:
+            url = (
+                os.environ.get("MOONSHOT_API_BASE_URL")
+                or "https://api.moonshot.ai/v1"
+            )
         timeout = timeout or float(os.environ.get("MODEL_TIMEOUT", 180))
         super().__init__(
             model_type=model_type,
@@ -81,27 +111,171 @@ class MoonshotModel(OpenAICompatibleModel):
             url=url,
             token_counter=token_counter,
             timeout=timeout,
+            max_retries=max_retries,
+            **kwargs,
         )
 
+    def _prepare_request(
+        self,
+        messages: List[OpenAIMessage],
+        response_format: Optional[Type[BaseModel]] = None,
+        tools: Optional[List[Dict[str, Any]]] = None,
+    ) -> Dict[str, Any]:
+        r"""Prepare the request configuration for Moonshot API.
+
+        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:
+            Dict[str, Any]: The prepared request configuration.
+        """
+        request_config = copy.deepcopy(self.model_config_dict)
+
+        if tools:
+            # Clean tools to remove null types (Moonshot API incompatibility)
+            cleaned_tools = self._clean_tool_schemas(tools)
+            request_config["tools"] = cleaned_tools
+        elif response_format:
+            # Use the same approach as DeepSeek for structured output
+            try_modify_message_with_format(messages[-1], response_format)
+            request_config["response_format"] = {"type": "json_object"}
+
+        return request_config
+
+    def _clean_tool_schemas(
+        self, tools: List[Dict[str, Any]]
+    ) -> List[Dict[str, Any]]:
+        r"""Clean tool schemas to remove null types for Moonshot compatibility.
+
+        Moonshot API doesn't accept {"type": "null"} in anyOf schemas.
+        This method removes null type definitions from parameters.
+
+        Args:
+            tools (List[Dict[str, Any]]): Original tool schemas.
+
+        Returns:
+            List[Dict[str, Any]]: Cleaned tool schemas.
+        """
+
+        def remove_null_from_schema(schema: Any) -> Any:
+            """Recursively remove null types from schema."""
+            if isinstance(schema, dict):
+                # Create a copy to avoid modifying the original
+                result = {}
+
+                for key, value in schema.items():
+                    if key == 'type' and isinstance(value, list):
+                        # Handle type arrays like ["string", "null"]
+                        filtered_types = [t for t in value if t != 'null']
+                        if len(filtered_types) == 1:
+                            # Single type remains, convert to string
+                            result[key] = filtered_types[0]
+                        elif len(filtered_types) > 1:
+                            # Multiple types remain, keep as array
+                            result[key] = filtered_types
+                        else:
+                            # All were null, use string as fallback
+                            logger.warning(
+                                "All types in tool schema type array "
+                                "were null, falling back to 'string' "
+                                "type for Moonshot API compatibility. "
+                                "Original tool schema may need review."
+                            )
+                            result[key] = 'string'
+                    elif key == 'anyOf':
+                        # Handle anyOf with null types
+                        filtered = [
+                            item
+                            for item in value
+                            if not (
+                                isinstance(item, dict)
+                                and item.get('type') == 'null'
+                            )
+                        ]
+                        if len(filtered) == 1:
+                            # If only one type remains, flatten it
+                            return remove_null_from_schema(filtered[0])
+                        elif len(filtered) > 1:
+                            result[key] = [
+                                remove_null_from_schema(item)
+                                for item in filtered
+                            ]
+                        else:
+                            # All were null, return string type as fallback
+                            logger.warning(
+                                "All types in tool schema anyOf were null, "
+                                "falling back to 'string' type for "
+                                "Moonshot API compatibility. Original "
+                                "tool schema may need review."
+                            )
+                            return {"type": "string"}
+                    else:
+                        # Recursively process other values
+                        result[key] = remove_null_from_schema(value)
+
+                return result
+            elif isinstance(schema, list):
+                return [remove_null_from_schema(item) for item in schema]
+            else:
+                return schema
+
+        cleaned_tools = copy.deepcopy(tools)
+        for tool in cleaned_tools:
+            if 'function' in tool and 'parameters' in tool['function']:
+                params = tool['function']['parameters']
+                if 'properties' in params:
+                    params['properties'] = remove_null_from_schema(
+                        params['properties']
+                    )
+
+        return cleaned_tools
+
+    @observe()
     async def _arun(
         self,
         messages: List[OpenAIMessage],
         response_format: Optional[Type[BaseModel]] = None,
         tools: Optional[List[Dict[str, Any]]] = None,
     ) -> Union[ChatCompletion, AsyncStream[ChatCompletionChunk]]:
-        raise NotImplementedError("Moonshot does not support async inference.")
+        r"""Runs inference of Moonshot chat completion asynchronously.
 
-    def check_model_config(self):
-        r"""Check whether the model configuration contains any
-        unexpected arguments to Moonshot API.
+        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.
 
-        Raises:
-            ValueError: If the model configuration dictionary contains any
-                unexpected arguments to Moonshot API.
+        Returns:
+            Union[ChatCompletion, AsyncStream[ChatCompletionChunk]]:
+                `ChatCompletion` in the non-stream mode, or
+                `AsyncStream[ChatCompletionChunk]` in the stream mode.
         """
-        for param in self.model_config_dict:
-            if param not in MOONSHOT_API_PARAMS:
-                raise ValueError(
-                    f"Unexpected argument `{param}` is "
-                    "input into Moonshot model backend."
-                )
+
+        # Update Langfuse trace with current agent session and metadata
+        agent_session_id = get_current_agent_session_id()
+        if agent_session_id:
+            update_langfuse_trace(
+                session_id=agent_session_id,
+                metadata={
+                    "agent_id": agent_session_id,
+                    "model_type": str(self.model_type),
+                },
+                tags=["CAMEL-AI", str(self.model_type)],
+            )
+
+        request_config = self._prepare_request(
+            messages, response_format, tools
+        )
+
+        return await self._async_client.chat.completions.create(
+            messages=messages,
+            model=self.model_type,
+            **request_config,
+        )

camel/models/nebius_model.py

@@ -0,0 +1,83 @@
+# ========= Copyright 2023-2025 @ 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 2023-2025 @ CAMEL-AI.org. All Rights Reserved. =========
+import os
+from typing import Any, Dict, Optional, Union
+
+from camel.configs import NebiusConfig
+from camel.models.openai_compatible_model import OpenAICompatibleModel
+from camel.types import ModelType
+from camel.utils import (
+    BaseTokenCounter,
+    api_keys_required,
+)
+
+
+class NebiusModel(OpenAICompatibleModel):
+    r"""LLM API served by Nebius AI Studio in a unified OpenAICompatibleModel
+    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:`NebiusConfig().as_dict()` will be used.
+            (default: :obj:`None`)
+        api_key (Optional[str], optional): The API key for authenticating
+            with the Nebius AI Studio service. (default: :obj:`None`).
+        url (Optional[str], optional): The url to the Nebius AI Studio 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`)
+        timeout (Optional[float], optional): The timeout value in seconds for
+            API calls. If not provided, will fall back to the MODEL_TIMEOUT
+            environment variable or default to 180 seconds.
+            (default: :obj:`None`)
+        max_retries (int, optional): Maximum number of retries for API calls.
+            (default: :obj:`3`)
+        **kwargs (Any): Additional arguments to pass to the client
+            initialization.
+    """
+
+    @api_keys_required([("api_key", "NEBIUS_API_KEY")])
+    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,
+        timeout: Optional[float] = None,
+        max_retries: int = 3,
+        **kwargs: Any,
+    ) -> None:
+        if model_config_dict is None:
+            model_config_dict = NebiusConfig().as_dict()
+        api_key = api_key or os.environ.get("NEBIUS_API_KEY")
+        url = url or os.environ.get(
+            "NEBIUS_API_BASE_URL", "https://api.studio.nebius.com/v1"
+        )
+        timeout = timeout or float(os.environ.get("MODEL_TIMEOUT", 180))
+        super().__init__(
+            model_type=model_type,
+            model_config_dict=model_config_dict,
+            api_key=api_key,
+            url=url,
+            token_counter=token_counter,
+            timeout=timeout,
+            max_retries=max_retries,
+            **kwargs,
+        )