camel-ai 0.2.44__py3-none-any.whl → 0.2.46__py3-none-any.whl

camel/models/ollama_model.py

@@ -13,24 +13,15 @@
 # ========= 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 typing import Any, Dict, Optional, Union
 
 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
+from camel.models.openai_compatible_model import OpenAICompatibleModel
+from camel.types import ModelType
+from camel.utils import BaseTokenCounter
 
 
-class OllamaModel(BaseModelBackend):
+class OllamaModel(OpenAICompatibleModel):
     r"""Ollama service interface.
 
     Args:
@@ -72,23 +63,16 @@ class OllamaModel(BaseModelBackend):
         url = url or os.environ.get("OLLAMA_BASE_URL")
         timeout = timeout or float(os.environ.get("MODEL_TIMEOUT", 180))
         super().__init__(
-            model_type, model_config_dict, api_key, url, token_counter, timeout
+            model_type=model_type,
+            model_config_dict=model_config_dict,
+            api_key=api_key,
+            url=url,
+            token_counter=token_counter,
+            timeout=timeout,
         )
+
         if not self._url:
             self._start_server()
-        # Use OpenAI client as interface call Ollama
-        self._client = OpenAI(
-            timeout=self._timeout,
-            max_retries=3,
-            api_key="Set-but-ignored",  # required but ignored
-            base_url=self._url,
-        )
-        self._async_client = AsyncOpenAI(
-            timeout=self._timeout,
-            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."""
@@ -106,18 +90,6 @@ class OllamaModel(BaseModelBackend):
         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
-
     def check_model_config(self):
         r"""Check whether the model configuration contains any
         unexpected arguments to Ollama API.
@@ -132,168 +104,3 @@ class OllamaModel(BaseModelBackend):
                     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:
-        import copy
-
-        request_config = copy.deepcopy(self.model_config_dict)
-        # Remove stream from request_config since Ollama does not support it
-        # when structured response is used
-        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:
-        import copy
-
-        request_config = copy.deepcopy(self.model_config_dict)
-        # Remove stream from request_config since Ollama does not support it
-        # when structured response is used
-        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)

camel/models/openai_compatible_model.py

@@ -13,13 +13,16 @@
 # ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
 
 import os
+from json import JSONDecodeError
 from typing import Any, Dict, List, Optional, Type, Union
 
 from openai import AsyncOpenAI, AsyncStream, OpenAI, Stream
-from pydantic import BaseModel
+from pydantic import BaseModel, ValidationError
 
+from camel.logger import get_logger
 from camel.messages import OpenAIMessage
-from camel.models import BaseModelBackend
+from camel.models._utils import try_modify_message_with_format
+from camel.models.base_model import BaseModelBackend
 from camel.types import (
     ChatCompletion,
     ChatCompletionChunk,
@@ -30,6 +33,8 @@ from camel.utils import (
     OpenAITokenCounter,
 )
 
+logger = get_logger(__name__)
+
 
 class OpenAICompatibleModel(BaseModelBackend):
     r"""Constructor for model backend supporting OpenAI compatibility.
@@ -187,11 +192,28 @@ class OpenAICompatibleModel(BaseModelBackend):
         if tools is not None:
             request_config["tools"] = tools
 
-        return self._client.beta.chat.completions.parse(
-            messages=messages,
-            model=self.model_type,
-            **request_config,
-        )
+        try:
+            return self._client.beta.chat.completions.parse(
+                messages=messages,
+                model=self.model_type,
+                **request_config,
+            )
+        except (ValidationError, JSONDecodeError) as e:
+            logger.warning(
+                f"Format validation error: {e}. "
+                f"Attempting fallback with JSON format."
+            )
+            try_modify_message_with_format(messages[-1], response_format)
+            request_config["response_format"] = {"type": "json_object"}
+            try:
+                return self._client.beta.chat.completions.parse(
+                    messages=messages,
+                    model=self.model_type,
+                    **request_config,
+                )
+            except Exception as e:
+                logger.error(f"Fallback attempt also failed: {e}")
+                raise
 
     async def _arequest_parse(
         self,
@@ -209,11 +231,28 @@ class OpenAICompatibleModel(BaseModelBackend):
         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,
-        )
+        try:
+            return await self._async_client.beta.chat.completions.parse(
+                messages=messages,
+                model=self.model_type,
+                **request_config,
+            )
+        except (ValidationError, JSONDecodeError) as e:
+            logger.warning(
+                f"Format validation error: {e}. "
+                f"Attempting fallback with JSON format."
+            )
+            try_modify_message_with_format(messages[-1], response_format)
+            request_config["response_format"] = {"type": "json_object"}
+            try:
+                return await self._async_client.beta.chat.completions.parse(
+                    messages=messages,
+                    model=self.model_type,
+                    **request_config,
+                )
+            except Exception as e:
+                logger.error(f"Fallback attempt also failed: {e}")
+                raise
 
     @property
     def token_counter(self) -> BaseTokenCounter:

camel/models/openrouter_model.py

@@ -12,29 +12,20 @@
 # limitations under the License.
 # ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
 import os
-from typing import Any, Dict, List, Optional, Type, Union
-
-from openai import AsyncOpenAI, AsyncStream, OpenAI, Stream
-from pydantic import BaseModel
+from typing import Any, Dict, Optional, Union
 
 from camel.configs import OPENROUTER_API_PARAMS, OpenRouterConfig
-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.models.openai_compatible_model import OpenAICompatibleModel
+from camel.types import ModelType
 from camel.utils import (
     BaseTokenCounter,
-    OpenAITokenCounter,
     api_keys_required,
 )
 
 
-class OpenRouterModel(BaseModelBackend):
-    r"""LLM API served by OpenRouter in a unified BaseModelBackend interface.
+class OpenRouterModel(OpenAICompatibleModel):
+    r"""LLM API served by OpenRouter in a unified OpenAICompatibleModel
+    interface.
 
     Args:
         model_type (Union[ModelType, str]): Model for which a backend is
@@ -75,114 +66,14 @@ class OpenRouterModel(BaseModelBackend):
         )
         timeout = timeout or float(os.environ.get("MODEL_TIMEOUT", 180))
         super().__init__(
-            model_type, model_config_dict, api_key, url, token_counter, timeout
-        )
-        self._client = OpenAI(
-            timeout=self._timeout,
-            max_retries=3,
-            api_key=self._api_key,
-            base_url=self._url,
-        )
-        self._async_client = AsyncOpenAI(
-            timeout=self._timeout,
-            max_retries=3,
-            api_key=self._api_key,
-            base_url=self._url,
-        )
-
-    @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
-
-    def _prepare_request(
-        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
-        elif response_format:
-            try_modify_message_with_format(messages[-1], response_format)
-            request_config["response_format"] = {"type": "json_object"}
-
-        return request_config
-
-    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 OpenAI 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.
-        """
-        request_config = self._prepare_request(
-            messages, response_format, tools
+            model_type=model_type,
+            model_config_dict=model_config_dict,
+            api_key=api_key,
+            url=url,
+            token_counter=token_counter,
+            timeout=timeout,
         )
 
-        response = self._client.chat.completions.create(
-            messages=messages,
-            model=self.model_type,
-            **request_config,
-        )
-
-        return response
-
-    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 OpenRouter chat completion asynchronously.
-
-        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.
-        """
-        request_config = self._prepare_request(
-            messages, response_format, tools
-        )
-
-        response = await self._async_client.chat.completions.create(
-            messages=messages,
-            model=self.model_type,
-            **request_config,
-        )
-
-        return response
-
     def check_model_config(self):
         r"""Check whether the model configuration contains any unexpected
         arguments to OpenRouter API. But OpenRouter API does not have any
@@ -198,13 +89,3 @@ class OpenRouterModel(BaseModelBackend):
                     f"Unexpected argument `{param}` is "
                     "input into OpenRouter 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/ppio_model.py

@@ -13,27 +13,18 @@
 # ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
 
 import os
-from typing import Any, Dict, List, Optional, Type, Union
-
-from openai import AsyncOpenAI, AsyncStream, OpenAI, Stream
-from pydantic import BaseModel
+from typing import Any, Dict, Optional, Union
 
 from camel.configs import PPIO_API_PARAMS, PPIOConfig
-from camel.messages import OpenAIMessage
-from camel.models import BaseModelBackend
-from camel.types import (
-    ChatCompletion,
-    ChatCompletionChunk,
-    ModelType,
-)
+from camel.models.openai_compatible_model import OpenAICompatibleModel
+from camel.types import ModelType
 from camel.utils import (
     BaseTokenCounter,
-    OpenAITokenCounter,
     api_keys_required,
 )
 
 
-class PPIOModel(BaseModelBackend):
+class PPIOModel(OpenAICompatibleModel):
     r"""Constructor for PPIO backend with OpenAI compatibility.
 
     Args:
@@ -79,85 +70,15 @@ class PPIOModel(BaseModelBackend):
             "PPIO_API_BASE_URL", "https://api.ppinfra.com/v3/openai"
         )
         timeout = timeout or float(os.environ.get("MODEL_TIMEOUT", 180))
-
         super().__init__(
-            model_type, model_config_dict, api_key, url, token_counter, timeout
+            model_type=model_type,
+            model_config_dict=model_config_dict,
+            api_key=api_key,
+            url=url,
+            token_counter=token_counter,
+            timeout=timeout,
         )
 
-        self._client = OpenAI(
-            timeout=self._timeout,
-            max_retries=3,
-            api_key=self._api_key,
-            base_url=self._url,
-        )
-        self._async_client = AsyncOpenAI(
-            timeout=self._timeout,
-            max_retries=3,
-            api_key=self._api_key,
-            base_url=self._url,
-        )
-
-    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 OpenAI chat completion.
-
-        Args:
-            messages (List[OpenAIMessage]): Message list with the chat history
-                in OpenAI API format.
-
-        Returns:
-            Union[ChatCompletion, AsyncStream[ChatCompletionChunk]]:
-                `ChatCompletion` in the non-stream mode, or
-                `AsyncStream[ChatCompletionChunk]` in the stream mode.
-        """
-        response = await self._async_client.chat.completions.create(
-            messages=messages,
-            model=self.model_type,
-            **self.model_config_dict,
-        )
-        return response
-
-    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 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_type,
-            **self.model_config_dict,
-        )
-        return response
-
-    @property
-    def token_counter(self) -> BaseTokenCounter:
-        r"""Initialize the token counter for the model backend.
-
-        Returns:
-            OpenAITokenCounter: 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
-
     def check_model_config(self):
         r"""Check whether the model configuration contains any
         unexpected arguments to PPIO API.
@@ -172,13 +93,3 @@ class PPIOModel(BaseModelBackend):
                     f"Unexpected argument `{param}` is "
                     "input into PPIO 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)