camel-ai 0.2.11__py3-none-any.whl → 0.2.12__py3-none-any.whl

camel/loaders/base_io.py

@@ -22,6 +22,47 @@ from typing import Any, Dict, List, Optional
 from camel.utils import dependencies_required
 
 
+def create_file(file: BytesIO, filename: str) -> "File":
+    r"""Reads an uploaded file and returns a File object.
+
+    Args:
+        file (BytesIO): A BytesIO object representing the contents of the
+            file.
+        filename (str): The name of the file.
+
+    Returns:
+        File: A File object.
+    """
+    ext_to_cls = {
+        "docx": DocxFile,
+        "pdf": PdfFile,
+        "txt": TxtFile,
+        "json": JsonFile,
+        "html": HtmlFile,
+    }
+
+    ext = filename.split(".")[-1].lower()
+    if ext not in ext_to_cls:
+        raise NotImplementedError(f"File type {ext} not supported")
+
+    out_file = ext_to_cls[ext].from_bytes(file, filename)
+    return out_file
+
+
+def create_file_from_raw_bytes(raw_bytes: bytes, filename: str) -> "File":
+    r"""Reads raw bytes and returns a File object.
+
+    Args:
+        raw_bytes (bytes): The raw bytes content of the file.
+        filename (str): The name of the file.
+
+    Returns:
+        File: A File object.
+    """
+    file = BytesIO(raw_bytes)
+    return create_file(file, filename)
+
+
 class File(ABC):
     r"""Represents an uploaded file comprised of Documents.
 
@@ -79,47 +120,6 @@ class File(ABC):
         file = BytesIO(raw_bytes)
         return cls.from_bytes(file, filename)
 
-    @staticmethod
-    def create_file(file: BytesIO, filename: str) -> "File":
-        r"""Reads an uploaded file and returns a File object.
-
-        Args:
-            file (BytesIO): A BytesIO object representing the contents of the
-                file.
-            filename (str): The name of the file.
-
-        Returns:
-            File: A File object.
-        """
-        ext_to_cls = {
-            "docx": DocxFile,
-            "pdf": PdfFile,
-            "txt": TxtFile,
-            "json": JsonFile,
-            "html": HtmlFile,
-        }
-
-        ext = filename.split(".")[-1].lower()
-        if ext not in ext_to_cls:
-            raise NotImplementedError(f"File type {ext} not supported")
-
-        out_file = ext_to_cls[ext].from_bytes(file, filename)
-        return out_file
-
-    @staticmethod
-    def create_file_from_raw_bytes(raw_bytes: bytes, filename: str) -> "File":
-        r"""Reads raw bytes and returns a File object.
-
-        Args:
-            raw_bytes (bytes): The raw bytes content of the file.
-            filename (str): The name of the file.
-
-        Returns:
-            File: A File object.
-        """
-        file = BytesIO(raw_bytes)
-        return File.create_file(file, filename)
-
     def __repr__(self) -> str:
         return (
             f"File(name={self.name}, id={self.file_id}, "

camel/messages/__init__.py

@@ -20,6 +20,7 @@ from camel.types import (
 )
 
 from .conversion import (
+    AlpacaItem,
     HermesFunctionFormatter,
     ShareGPTMessage,
 )
@@ -52,4 +53,5 @@ __all__ = [
     'ShareGPTMessage',
     'BaseMessage',
     'FunctionCallingMessage',
+    'AlpacaItem',
 ]

camel/models/__init__.py

@@ -31,6 +31,7 @@ from .openai_model import OpenAIModel
 from .qwen_model import QwenModel
 from .reka_model import RekaModel
 from .samba_model import SambaModel
+from .sglang_model import SGLangModel
 from .stub_model import StubModel
 from .togetherai_model import TogetherAIModel
 from .vllm_model import VLLMModel
@@ -55,6 +56,7 @@ __all__ = [
     'NvidiaModel',
     'OllamaModel',
     'VLLMModel',
+    'SGLangModel',
     'GeminiModel',
     'OpenAICompatibleModel',
     'RekaModel',

camel/models/anthropic_model.py

@@ -12,7 +12,7 @@
 # limitations under the License.
 # ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
 import os
-from typing import Any, Dict, List, Optional, Union
+from typing import Any, Dict, List, Literal, Optional, Union
 
 from camel.configs import ANTHROPIC_API_PARAMS, AnthropicConfig
 from camel.messages import OpenAIMessage
@@ -94,19 +94,29 @@ class AnthropicModel(BaseModelBackend):
                 tokenization style.
         """
         if not self._token_counter:
-            self._token_counter = AnthropicTokenCounter()
+            self._token_counter = AnthropicTokenCounter(self.model_type)
         return self._token_counter
 
-    def count_tokens_from_prompt(self, prompt: str) -> int:
+    @dependencies_required('anthropic')
+    def count_tokens_from_prompt(
+        self, prompt: str, role: Literal["user", "assistant"]
+    ) -> int:
         r"""Count the number of tokens from a prompt.
 
         Args:
             prompt (str): The prompt string.
+            role (Literal["user", "assistant"]): The role of the message
+                sender, either "user" or "assistant".
 
         Returns:
             int: The number of tokens in the prompt.
         """
-        return self.client.count_tokens(prompt)
+        from anthropic.types.beta import BetaMessageParam
+
+        return self.client.beta.messages.count_tokens(
+            messages=[BetaMessageParam(content=prompt, role=role)],
+            model=self.model_type,
+        ).input_tokens
 
     @api_keys_required("ANTHROPIC_API_KEY")
     def run(

camel/models/base_model.py

@@ -21,6 +21,7 @@ from camel.types import (
     ChatCompletion,
     ChatCompletionChunk,
     ModelType,
+    ParsedChatCompletion,
     UnifiedModelType,
 )
 from camel.utils import BaseTokenCounter
@@ -114,6 +115,33 @@ class BaseModelBackend(ABC):
         """
         return self.token_counter.count_tokens_from_messages(messages)
 
+    def _to_chat_completion(
+        self, response: ParsedChatCompletion
+    ) -> ChatCompletion:
+        if len(response.choices) > 1:
+            print("Warning: Multiple response choices detected")
+
+        choice = dict(
+            index=response.choices[0].index,
+            message={
+                "role": response.choices[0].message.role,
+                "content": response.choices[0].message.content,
+                "tool_calls": response.choices[0].message.tool_calls,
+                "parsed": response.choices[0].message.parsed,
+            },
+            finish_reason=response.choices[0].finish_reason,
+        )
+
+        obj = ChatCompletion.construct(
+            id=response.id,
+            choices=[choice],
+            created=response.created,
+            model=response.model,
+            object="chat.completion",
+            usage=response.usage,
+        )
+        return obj
+
     @property
     def token_limit(self) -> int:
         r"""Returns the maximum token limit for a given model.

camel/models/groq_model.py

@@ -63,7 +63,7 @@ class GroqModel(BaseModelBackend):
             model_config_dict = GroqConfig().as_dict()
         api_key = api_key or os.environ.get("GROQ_API_KEY")
         url = url or os.environ.get(
-            "GROQ_API_BASE_URL" or "https://api.groq.com/openai/v1"
+            "GROQ_API_BASE_URL", "https://api.groq.com/openai/v1"
         )
         super().__init__(
             model_type, model_config_dict, api_key, url, token_counter

camel/models/model_factory.py

@@ -29,6 +29,7 @@ from camel.models.openai_model import OpenAIModel
 from camel.models.qwen_model import QwenModel
 from camel.models.reka_model import RekaModel
 from camel.models.samba_model import SambaModel
+from camel.models.sglang_model import SGLangModel
 from camel.models.stub_model import StubModel
 from camel.models.togetherai_model import TogetherAIModel
 from camel.models.vllm_model import VLLMModel
@@ -86,6 +87,8 @@ class ModelFactory:
             model_class = OllamaModel
         elif model_platform.is_vllm:
             model_class = VLLMModel
+        elif model_platform.is_sglang:
+            model_class = SGLangModel
         elif model_platform.is_openai_compatible_model:
             model_class = OpenAICompatibleModel
         elif model_platform.is_samba:

camel/models/ollama_model.py

@@ -134,6 +134,18 @@ class OllamaModel(BaseModelBackend):
                 `ChatCompletion` in the non-stream mode, or
                 `Stream[ChatCompletionChunk]` in the stream mode.
         """
+        if self.model_config_dict.get("response_format"):
+            # stream is not supported in beta.chat.completions.parse
+            if "stream" in self.model_config_dict:
+                del self.model_config_dict["stream"]
+
+            response = self._client.beta.chat.completions.parse(
+                messages=messages,
+                model=self.model_type,
+                **self.model_config_dict,
+            )
+
+            return self._to_chat_completion(response)
 
         response = self._client.chat.completions.create(
             messages=messages,

camel/models/openai_model.py

@@ -24,7 +24,6 @@ from camel.types import (
     ChatCompletion,
     ChatCompletionChunk,
     ModelType,
-    ParsedChatCompletion,
 )
 from camel.utils import (
     BaseTokenCounter,
@@ -148,31 +147,6 @@ class OpenAIModel(BaseModelBackend):
         )
         return response
 
-    def _to_chat_completion(
-        self, response: "ParsedChatCompletion"
-    ) -> ChatCompletion:
-        # TODO: Handle n > 1 or warn consumers it's not supported
-        choice = dict(
-            index=response.choices[0].index,
-            message={
-                "role": response.choices[0].message.role,
-                "content": response.choices[0].message.content,
-                "tool_calls": response.choices[0].message.tool_calls,
-                "parsed": response.choices[0].message.parsed,
-            },
-            finish_reason=response.choices[0].finish_reason,
-        )
-
-        obj = ChatCompletion.construct(
-            id=response.id,
-            choices=[choice],
-            created=response.created,
-            model=response.model,
-            object="chat.completion",
-            usage=response.usage,
-        )
-        return obj
-
     def check_model_config(self):
         r"""Check whether the model configuration contains any
         unexpected arguments to OpenAI API.

camel/models/reward/__init__.py

@@ -0,0 +1,22 @@
+# ========= 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. =========
+from .base_reward_model import BaseRewardModel
+from .evaluator import Evaluator
+from .nemotron_model import NemotronRewardModel
+
+__all__ = [
+    'BaseRewardModel',
+    'NemotronRewardModel',
+    'Evaluator',
+]

camel/models/reward/base_reward_model.py

@@ -0,0 +1,58 @@
+# ========= 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. =========
+from abc import ABC, abstractmethod
+from typing import Dict, List, Optional, Union
+
+from camel.types import ModelType
+
+
+class BaseRewardModel(ABC):
+    r"""Abstract base class for reward models. Reward models are used to
+    evaluate messages and return scores based on different criteria.
+
+    Subclasses should implement the 'evaluate' and 'get_scores_types' methods.
+    """
+
+    def __init__(
+        self,
+        model_type: Union[ModelType, str],
+        api_key: Optional[str] = None,
+        url: Optional[str] = None,
+    ) -> None:
+        self.model_type = model_type
+        self.api_key = api_key
+        self.url = url
+
+    @abstractmethod
+    def evaluate(self, messages: List[Dict[str, str]]) -> Dict[str, float]:
+        r"""Evaluate the messages and return scores based on different
+        criteria.
+
+        Args:
+            messages (List[Dict[str, str]]): A list of messages where each
+                message is a dictionary with 'role' and 'content'.
+
+        Returns:
+            Dict[str, float]: A dictionary mapping score types to their values.
+        """
+        pass
+
+    @abstractmethod
+    def get_scores_types(self) -> List[str]:
+        r"""Get the list of score types that the reward model can return.
+
+        Returns:
+            List[str]: A list of score types that the reward model can return.
+        """
+        pass

camel/models/reward/evaluator.py

@@ -0,0 +1,63 @@
+# ========= 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. =========
+from typing import Dict, List
+
+from camel.models.reward import BaseRewardModel
+
+
+class Evaluator:
+    r"""Evaluator class to evaluate messages using a reward model and filter
+    data based on the scores.
+
+    Args:
+        reward_model (BaseRewardModel): A reward model to evaluate messages.
+    """
+
+    def __init__(self, reward_model: BaseRewardModel):
+        self.reward_model = reward_model
+
+    def evaluate(self, messages: List[Dict[str, str]]) -> Dict[str, float]:
+        r"""Evaluate the messages using the reward model.
+
+        Args:
+            messages (List[Dict[str, str]]): A list of messages where each
+                message is a dictionary with 'role' and 'content'.
+
+        Returns:
+            Dict[str, float]: A dictionary mapping score types to their values.
+        """
+        scores = self.reward_model.evaluate(messages)
+        return scores
+
+    def filter_data(
+        self, messages: List[Dict[str, str]], thresholds: Dict[str, float]
+    ) -> bool:
+        r"""Filter messages based on the scores.
+
+        Args:
+            messages (List[Dict[str, str]]): A list of messages where each
+                message is a dictionary with 'role' and 'content'.
+            thresholds (Dict[str, float]): A dictionary mapping score types to
+                their values.
+
+        Returns:
+            bool: A boolean indicating whether the messages pass the filter.
+        """
+        scores = self.evaluate(messages)
+        for score_type, threshold in thresholds.items():
+            if score_type not in scores:
+                raise ValueError(f"Score type {score_type} not found.")
+            if scores.get(score_type, 0) < threshold:
+                return False
+        return True

camel/models/reward/nemotron_model.py

@@ -0,0 +1,112 @@
+# ========= 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
+from typing import Dict, List, Optional, Union
+
+from openai import OpenAI
+
+from camel.models.reward import BaseRewardModel
+from camel.types import ChatCompletion, ModelType
+from camel.utils import api_keys_required
+
+
+class NemotronRewardModel(BaseRewardModel):
+    r"""Reward model based on the Nemotron model with OpenAI compatibility.
+
+    Args:
+        model_type (Union[ModelType, str]): Model for which a backend is
+            created.
+        api_key (Optional[str], optional): The API key for authenticating
+            with the model service. (default: :obj:`None`)
+        url (Optional[str], optional): The url to the model service.
+
+    Note:
+        The Nemotron model does not support model config.
+    """
+
+    def __init__(
+        self,
+        model_type: Union[ModelType, str],
+        api_key: Optional[str] = None,
+        url: Optional[str] = None,
+    ) -> None:
+        url = url or os.environ.get(
+            "NVIDIA_API_BASE_URL", "https://integrate.api.nvidia.com/v1"
+        )
+        api_key = api_key or os.environ.get("NVIDIA_API_KEY")
+        super().__init__(model_type, api_key, url)
+        self._client = OpenAI(
+            timeout=60,
+            max_retries=3,
+            base_url=self.url,
+            api_key=self.api_key,
+        )
+
+    @api_keys_required("NVIDIA_API_KEY")
+    def evaluate(self, messages: List[Dict[str, str]]) -> Dict[str, float]:
+        r"""Evaluate the messages using the Nemotron model.
+
+        Args:
+            messages (List[Dict[str, str]]): A list of messages where each
+                message is a dictionary format.
+
+        Returns:
+            Dict[str, float]:  A dictionary mapping score types to their
+                values.
+        """
+        response = self._client.chat.completions.create(
+            messages=messages,  # type: ignore[arg-type]
+            model=self.model_type,
+        )
+        scores = self._parse_scores(response)
+        return scores
+
+    def get_scores_types(self) -> List[str]:
+        r"""Get the list of score types that the reward model can return.
+
+        Returns:
+            List[str]: A list of score types that the reward model can return.
+        """
+        return [
+            "helpfulness",
+            "correctness",
+            "coherence",
+            "complexity",
+            "verbosity",
+        ]
+
+    def _parse_scores(self, response: ChatCompletion) -> Dict[str, float]:
+        r"""Parse the scores from the response.
+
+        Args:
+            response (ChatCompletion): A ChatCompletion object with the scores.
+
+        Returns:
+            Dict[str, float]: A dictionary mapping score types to their values.
+        """
+        try:
+            choices = response.choices
+            logprobs = (
+                choices[0].logprobs.content
+                if choices and choices[0].logprobs
+                else None
+            )
+            scores = (
+                {entry.token: entry.logprob for entry in logprobs if entry}
+                if logprobs
+                else {}
+            )
+            return scores
+        except Exception as e:
+            raise ValueError(f"Failed to parse scores: {e}")