Skip to content

armgit / camel

Public Imported from GitHub
0

🐫 CAMEL: The first and the best multi-agent framework. Finding the Scaling Law of Agents. https://www.camel-ai.org

fix: Remove <think> content in model request (#1613)

Wendong-Fan457e7c5dffc8
diff --git a/.github/ISSUE_TEMPLATE/bug_report.yml b/.github/ISSUE_TEMPLATE/bug_report.yml
index 75924e689..227c4e974 100644
--- a/.github/ISSUE_TEMPLATE/bug_report.yml
+++ b/.github/ISSUE_TEMPLATE/bug_report.yml
@@ -26,7 +26,7 @@ body:
attributes:
label: What version of camel are you using?
description: Run command `python3 -c 'print(__import__("camel").__version__)'` in your shell and paste the output here.
- placeholder: E.g., 0.2.21
+ placeholder: E.g., 0.2.22
validations:
required: true
diff --git a/camel/__init__.py b/camel/__init__.py
index 77b1760a7..8cb8df7f0 100644
--- a/camel/__init__.py
+++ b/camel/__init__.py
@@ -14,7 +14,7 @@
from camel.logger import disable_logging, enable_logging, set_log_level
-__version__ = '0.2.21'
+__version__ = '0.2.22'
__all__ = [
'__version__',
diff --git a/camel/models/base_model.py b/camel/models/base_model.py
index 37e9d4061..44ae64720 100644
--- a/camel/models/base_model.py
+++ b/camel/models/base_model.py
@@ -11,6 +11,8 @@
# See the License for the specific language governing permissions and
# limitations under the License.
# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. =========
+import abc
+import re
from abc import ABC, abstractmethod
from typing import Any, Dict, List, Optional, Union
@@ -27,7 +29,30 @@ from camel.types import (
from camel.utils import BaseTokenCounter
-class BaseModelBackend(ABC):
+class ModelBackendMeta(abc.ABCMeta):
+ r"""Metaclass that automatically preprocesses messages in run method.
+
+ Automatically wraps the run method of any class inheriting from
+ BaseModelBackend to preprocess messages (remove <think> tags) before they
+ are sent to the model.
+ """
+
+ def __new__(mcs, name, bases, namespace):
+ r"""Wraps run method with preprocessing if it exists in the class."""
+ if 'run' in namespace:
+ original_run = namespace['run']
+
+ def wrapped_run(
+ self, messages: List[OpenAIMessage], *args, **kwargs
+ ):
+ messages = self.preprocess_messages(messages)
+ return original_run(self, messages, *args, **kwargs)
+
+ namespace['run'] = wrapped_run
+ return super().__new__(mcs, name, bases, namespace)
+
+
+class BaseModelBackend(ABC, metaclass=ModelBackendMeta):
r"""Base class for different model backends.
It may be OpenAI API, a local LLM, a stub for unit tests, etc.
@@ -73,6 +98,34 @@ class BaseModelBackend(ABC):
"""
pass
+ def preprocess_messages(
+ self, messages: List[OpenAIMessage]
+ ) -> List[OpenAIMessage]:
+ r"""Preprocess messages before sending to model API.
+ Removes thinking content and other model-specific preprocessing.
+
+ Args:
+ messages (List[OpenAIMessage]): Original messages
+
+ Returns:
+ List[OpenAIMessage]: Preprocessed messages
+ """
+ # Remove thinking content from messages before sending to API
+ # This ensures only the final response is sent, excluding
+ # intermediate thought processes
+ return [
+ { # type: ignore[misc]
+ **msg,
+ 'content': re.sub(
+ r'<think>.*?</think>',
+ '',
+ msg['content'], # type: ignore[arg-type]
+ flags=re.DOTALL,
+ ).strip(),
+ }
+ for msg in messages
+ ]
+
@abstractmethod
def run(
self,
diff --git a/camel/models/deepseek_model.py b/camel/models/deepseek_model.py
index 40375749f..56b694742 100644
--- a/camel/models/deepseek_model.py
+++ b/camel/models/deepseek_model.py
@@ -118,8 +118,6 @@ class DeepSeekModel(BaseModelBackend):
if self.model_type in [
ModelType.DEEPSEEK_REASONER,
]:
- import re
-
logger.warning(
"You are using a DeepSeek Reasoner model, "
"which has certain limitations, reference: "
@@ -141,22 +139,6 @@ class DeepSeekModel(BaseModelBackend):
if key in self.model_config_dict:
del self.model_config_dict[key]
- # Remove thinking content from messages before sending to API
- # This ensures only the final response is sent, excluding
- # intermediate thought processes
- messages = [
- { # type: ignore[misc]
- **msg,
- 'content': re.sub(
- r'<think>.*?</think>',
- '',
- msg['content'], # type: ignore[arg-type]
- flags=re.DOTALL,
- ).strip(),
- }
- for msg in messages
- ]
-
response = self._client.chat.completions.create(
messages=messages,
model=self.model_type,
diff --git a/docs/conf.py b/docs/conf.py
index f767aa6e3..fdf039a8f 100644
--- a/docs/conf.py
+++ b/docs/conf.py
@@ -27,7 +27,7 @@ sys.path.insert(0, os.path.abspath('..'))
project = 'CAMEL'
copyright = '2024, CAMEL-AI.org'
author = 'CAMEL-AI.org'
-release = '0.2.21'
+release = '0.2.22'
html_favicon = (
'https://raw.githubusercontent.com/camel-ai/camel/master/misc/favicon.png'
diff --git a/docs/get_started/installation.md b/docs/get_started/installation.md
index a978ca460..c85311ec3 100644
--- a/docs/get_started/installation.md
+++ b/docs/get_started/installation.md
@@ -60,7 +60,7 @@ conda create --name camel python=3.10
conda activate camel
# Clone github repo
-git clone -b v0.2.21 https://github.com/camel-ai/camel.git
+git clone -b v0.2.22 https://github.com/camel-ai/camel.git
# Change directory into project directory
cd camel
diff --git a/docs/key_modules/loaders.md b/docs/key_modules/loaders.md
index 0a74d9614..371a70c7d 100644
--- a/docs/key_modules/loaders.md
+++ b/docs/key_modules/loaders.md
@@ -340,14 +340,14 @@ response = jina_reader.read_content("https://docs.camel-ai.org/")
print(response)
```
```markdown
->>>Welcome to CAMEL’s documentation! — CAMEL 0.2.21 documentation
+>>>Welcome to CAMEL’s documentation! — CAMEL 0.2.22 documentation
===============
[Skip to main content](https://docs.camel-ai.org/#main-content)
Back to top Ctrl+K
- [![Image 1](https://raw.githubusercontent.com/camel-ai/camel/master/misc/logo_light.png) ![Image 2](https://raw.githubusercontent.com/camel-ai/camel/master/misc/logo_light.png)CAMEL 0.2.21](https://docs.camel-ai.org/#)
+ [![Image 1](https://raw.githubusercontent.com/camel-ai/camel/master/misc/logo_light.png) ![Image 2](https://raw.githubusercontent.com/camel-ai/camel/master/misc/logo_light.png)CAMEL 0.2.22](https://docs.camel-ai.org/#)
Search Ctrl+K
diff --git a/pyproject.toml b/pyproject.toml
index 5d6d679fe..7f42d885d 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
[tool.poetry]
name = "camel-ai"
-version = "0.2.21"
+version = "0.2.22"
authors = ["CAMEL-AI.org"]
description = "Communicative Agents for AI Society Study"
readme = "README.md"
diff --git a/test/models/test_base_model.py b/test/models/test_base_model.py
new file mode 100644
index 000000000..f84b7b1e1
--- /dev/null
+++ b/test/models/test_base_model.py
@@ -0,0 +1,89 @@
+# ========= 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 camel.models import BaseModelBackend
+from camel.types import ModelType
+
+
+class TestBaseModelBackend:
+ r"""Unit tests for the BaseModelBackend class."""
+
+ def test_preprocess_messages(self):
+ r"""Test message preprocessing removes thinking content correctly."""
+
+ class DummyModel(BaseModelBackend):
+ @property
+ def token_counter(self):
+ pass
+
+ def run(self, messages):
+ pass
+
+ def check_model_config(self):
+ pass
+
+ model = DummyModel(ModelType.GPT_4O_MINI)
+
+ # Test basic thinking removal
+ messages = [
+ {
+ 'role': 'user',
+ 'content': 'Hello <think>thinking about response</think> '
+ 'world',
+ },
+ {
+ 'role': 'assistant',
+ 'content': '<think>Let me think...\nThinking more...</'
+ 'think>Response',
+ },
+ ]
+
+ processed = model.preprocess_messages(messages)
+ assert len(processed) == 2
+ assert processed[0]['content'] == 'Hello world'
+ assert processed[1]['content'] == 'Response'
+
+ # Test message without thinking tags
+ messages = [{'role': 'user', 'content': 'plain message'}]
+ processed = model.preprocess_messages(messages)
+ assert processed[0]['content'] == 'plain message'
+
+ def test_metaclass_preprocessing(self):
+ r"""Test that metaclass automatically preprocesses messages in run
+ method."""
+ processed_messages = None
+
+ class TestModel(BaseModelBackend):
+ @property
+ def token_counter(self):
+ pass
+
+ def run(self, messages):
+ nonlocal processed_messages
+ processed_messages = messages
+ return None
+
+ def check_model_config(self):
+ pass
+
+ model = TestModel(ModelType.GPT_4O_MINI)
+ messages = [
+ {'role': 'user', 'content': 'Hello <think>hi</think> world'}
+ ]
+
+ # Call run method and verify messages were preprocessed
+ model.run(messages)
+ assert processed_messages is not None
+ assert processed_messages[0]['content'] == 'Hello world'