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

feat: add data collector for dataset generation (#1193)

Xukun Liu33c2787be077
diff --git a/camel/data_collector/__init__.py b/camel/data_collector/__init__.py
new file mode 100644
index 000000000..b209e7588
--- /dev/null
+++ b/camel/data_collector/__init__.py
@@ -0,0 +1,19 @@
+# ========= 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 .alpaca_collector import AlpacaDataCollector
+from .base import BaseDataCollector
+from .sharegpt_collector import ShareGPTDataCollector
+
+__all__ = ["BaseDataCollector", "AlpacaDataCollector", "ShareGPTDataCollector"]
diff --git a/camel/data_collector/alpaca_collector.py b/camel/data_collector/alpaca_collector.py
new file mode 100644
index 000000000..bfea503a7
--- /dev/null
+++ b/camel/data_collector/alpaca_collector.py
@@ -0,0 +1,127 @@
+# ========= 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 Any, Dict, List, Optional, Union
+
+from typing_extensions import Self
+
+from camel.agents import ChatAgent
+from camel.data_collector.base import BaseDataCollector
+from camel.messages import AlpacaItem, BaseMessage
+from camel.schemas import OpenAISchemaConverter
+
+# ruff: noqa: E501
+DEFAULT_CONVERTER_PROMPTS = """
+ Extract key entities and attributes from the conversations
+ and convert them into a structured JSON format.
+ For example:
+ Instruction: You are a helpful assistant.
+ User: When is the release date of the video game Portal?
+ Assistant: The release date of the video game Portal is October 9.
+ Your output should be:
+ {
+ "instruction": "You are a helpful assistant. When is the release date of the video game Portal?",
+ "input": "",
+ "output": "The release date of the video game Portal is October 9."
+ }
+"""
+
+
+class AlpacaDataCollector(BaseDataCollector):
+ def __init__(self) -> None:
+ super().__init__()
+ self.system_message: Optional[BaseMessage] = None
+ self.agent_name: Optional[str] = None
+
+ def record(
+ self,
+ agent: Union[List[ChatAgent], ChatAgent],
+ ) -> Self:
+ r"""Inject an agent into the data collector.
+
+ Args:
+ agent (Union[List[ChatAgent], ChatAgent]):
+ The agent to inject.
+ """
+ if not self.agent_name:
+ _agent = agent if isinstance(agent, ChatAgent) else agent[0]
+ self.agent_name = _agent.role_name
+ self.system_message = _agent._system_message
+ super().record(agent)
+ return self
+
+ def convert(self) -> Dict[str, Any]:
+ r"""Convert the collected data into a dictionary."""
+ if self.agent_name is None:
+ raise ValueError("No agent injected")
+
+ history = self.get_agent_history(self.agent_name)
+ if not history:
+ raise ValueError("No data collected.")
+
+ # Validate and process history
+ if len(history) == 3 and history[0].role == "system":
+ history = history[1:] # Ignore the system message.
+ elif len(history) != 2:
+ raise ValueError(
+ f"AlpacaDataCollector only supports one message pair, but "
+ f"got {len(history)}"
+ )
+
+ input_message, output_message = history
+ instruction = (
+ self.system_message.content if self.system_message else ""
+ ) + str(input_message.message)
+
+ data = {
+ "instruction": instruction,
+ "input": "",
+ "output": output_message.message,
+ }
+ self.data.append(data)
+ return data
+
+ def llm_convert(
+ self,
+ converter: Optional[OpenAISchemaConverter] = None,
+ prompt: Optional[str] = None,
+ ) -> Dict[str, str]:
+ r"""Convert collected data using an LLM schema converter.
+
+ Args:
+ converter (Optional[OpenAISchemaConverter], optional):
+ The converter to use. (default: :obj:`OpenAISchemaConverter`)
+ prompt (Optional[str], optional): Prompt to guide the conversion.
+ (default: :obj:`DEFAULT_CONVERTER_PROMPTS`)
+
+ Returns:
+ Dict[str, str]: The converted data.
+
+ Raises:
+ ValueError: If no agent is injected or data cannot be collected.
+ """
+ prompt = prompt or DEFAULT_CONVERTER_PROMPTS
+ converter = converter or OpenAISchemaConverter()
+
+ system = self.system_message.content if self.system_message else ""
+ context = [f"Instruction: {system}\n"]
+
+ for message in self.get_agent_history(str(self.agent_name)):
+ if message.role == "user":
+ context.append(f"User: {message.message}\n")
+ else:
+ context.append(f"{message.name}: {message.message}\n")
+ return converter.convert(
+ "\n".join(context), AlpacaItem, prompt=prompt
+ ).model_dump()
diff --git a/camel/data_collector/base.py b/camel/data_collector/base.py
new file mode 100644
index 000000000..f9a919a4c
--- /dev/null
+++ b/camel/data_collector/base.py
@@ -0,0 +1,211 @@
+# ========= 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 uuid
+from abc import ABC, abstractmethod
+from typing import Any, Dict, List, Literal, Optional, Tuple, Union
+from uuid import UUID
+
+from typing_extensions import Self
+
+from camel.agents import ChatAgent
+
+
+class CollectorData:
+ def __init__(
+ self,
+ id: UUID,
+ name: str,
+ role: Literal["user", "assistant", "system", "function"],
+ message: Optional[str] = None,
+ function_call: Optional[Dict[str, Any]] = None,
+ ) -> None:
+ r"""Create a data item store information about a message.
+ Used by the data collector.
+
+ Args:
+
+ id (UUID): The id of the message.
+ name (str): The name of the agent.
+ role (Literal["user", "assistant", "system", "function"]):
+ The role of the message.
+ message (Optional[str], optional): The message.
+ (default: :obj:`None`)
+ function_call (Optional[Dict[str, Any]], optional):
+ The function call. (default: :obj:`None`)
+
+ Raises:
+
+ ValueError: If the role is not supported.
+ ValueError: If the role is system and function call is provided.
+ ValueError: If neither message nor function call is provided.
+
+ """
+ if role not in ["user", "assistant", "system", "function"]:
+ raise ValueError(f"Role {role} not supported")
+ if role == "system" and function_call:
+ raise ValueError("System role cannot have function call")
+ if not message and not function_call:
+ raise ValueError(
+ "Either message or function call must be provided"
+ )
+ self.id = id
+ self.name = name
+ self.role = role
+ self.message = message
+ self.function_call = function_call
+
+ @staticmethod
+ def from_context(name, context: Dict[str, Any]) -> "CollectorData":
+ r"""Create a data collector from a context.
+
+ Args:
+ name (str): The name of the agent.
+ context (Dict[str, Any]): The context.
+
+ Returns:
+ CollectorData: The data collector.
+ """
+ return CollectorData(
+ id=uuid.uuid4(),
+ name=name,
+ role=context["role"],
+ message=context["content"],
+ function_call=context.get("function_call", None),
+ )
+
+
+class BaseDataCollector(ABC):
+ r"""Base class for data collectors."""
+
+ def __init__(self) -> None:
+ r"""Create a data collector."""
+ self.history: List[CollectorData] = []
+ self._recording = False
+ self.agents: List[Tuple[str, ChatAgent]] = []
+ self.data: List[Dict[str, Any]] = []
+
+ def step(
+ self,
+ role: Literal["user", "assistant", "system", "function"],
+ name: Optional[str] = None,
+ message: Optional[str] = None,
+ function_call: Optional[Dict[str, Any]] = None,
+ ) -> Self:
+ r"""Record a message.
+
+ Args:
+ role (Literal["user", "assistant", "system", "function"]):
+ The role of the message.
+ name (Optional[str], optional): The name of the agent.
+ (default: :obj:`None`)
+ message (Optional[str], optional): The message to record.
+ (default: :obj:`None`)
+ function_call (Optional[Dict[str, Any]], optional):
+ The function call to record. (default: :obj:`None`)
+
+ Returns:
+ Self: The data collector.
+
+ """
+
+ name = name or role
+
+ self.history.append(
+ CollectorData(
+ id=uuid.uuid4(),
+ name=name,
+ role=role,
+ message=message,
+ function_call=function_call,
+ )
+ )
+ return self
+
+ def record(
+ self,
+ agent: Union[List[ChatAgent], ChatAgent],
+ ) -> Self:
+ r"""Record agents.
+
+ Args:
+ agent (Union[List[ChatAgent], ChatAgent]):
+ The agent(s) to inject.
+ """
+ if not isinstance(agent, list):
+ agent = [agent]
+ for a in agent:
+ name = a.role_name
+ if not name:
+ name = f"{a.__class__.__name__}_{len(self.agents)}"
+ if name in [n for n, _ in self.agents]:
+ raise ValueError(f"Name {name} already exists")
+
+ self.agents.append((name, a))
+ return self
+
+ def start(self) -> Self:
+ r"""Start recording."""
+ self._recording = True
+ return self
+
+ def stop(self) -> Self:
+ r"""Stop recording."""
+ self._recording = False
+ return self
+
+ @property
+ def recording(self) -> bool:
+ r"""Whether the collector is recording."""
+ return self._recording
+
+ def reset(self, reset_agents: bool = True):
+ r"""Reset the collector.
+
+ Args:
+ reset_agents (bool, optional):
+ Whether to reset the agents. Defaults to True.
+ """
+ self.history = []
+ if reset_agents:
+ for _, agent in self.agents:
+ agent.reset()
+
+ @abstractmethod
+ def convert(self) -> Any:
+ r"""Convert the collected data."""
+ pass
+
+ @abstractmethod
+ def llm_convert(self, converter: Any, prompt: Optional[str] = None) -> Any:
+ r"""Convert the collected data."""
+ pass
+
+ def get_agent_history(self, name: str) -> List[CollectorData]:
+ r"""Get the message history of an agent.
+
+ Args:
+ name (str): The name of the agent.
+
+ Returns:
+ List[CollectorData]: The message history of the agent
+ """
+ if not self.history:
+ for _name, agent in self.agents:
+ if _name == name:
+ return [
+ CollectorData.from_context(name, dict(i))
+ for i in agent.memory.get_context()[0]
+ ]
+ return [msg for msg in self.history if msg.name == name]
diff --git a/camel/data_collector/sharegpt_collector.py b/camel/data_collector/sharegpt_collector.py
new file mode 100644
index 000000000..ff6b5ce99
--- /dev/null
+++ b/camel/data_collector/sharegpt_collector.py
@@ -0,0 +1,205 @@
+# ========= 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 json
+from typing import Any, ClassVar, Dict, List, Literal, Optional, Union
+
+from pydantic import BaseModel
+from typing_extensions import Self
+
+from camel.agents import ChatAgent
+from camel.data_collector.base import BaseDataCollector
+from camel.messages import BaseMessage
+from camel.messages.conversion.conversation_models import (
+ ShareGPTConversation,
+ ShareGPTMessage,
+)
+from camel.schemas import OpenAISchemaConverter
+from camel.toolkits import FunctionTool
+
+FROM_HASH = {
+ "human": "human",
+ "gpt": "gpt",
+ "observation": "human",
+ "function_call": "gpt",
+}
+# ruff: noqa: E501
+DEFAULT_CONVERTER_PROMPTS = """
+ Extract key entities and attributes from the conversations
+ and convert them into a structured JSON format.
+ For example:
+ System: You are a helpful assistant
+ Tools: [{"name": "get_release_date", "arguments": ["Portal"]}]
+ User: When is the release date of the video game Portal?
+ Assistant: The release date of the video game Portal is October 9, 2007.
+ Your output should be:
+ {
+ "system": "You are a helpful assistant",
+ "tools": "[{"name": "get_release_date", "arguments": ["Portal"]}]",
+ "conversations": [
+ {"from": "human", "value": "When is the release date of the video game Portal?"},
+ {"from": "gpt", "value": "The release date of the video game Portal is October 9, 2007."}
+ ]
+ }
+"""
+
+
+class ConversationItem(BaseModel):
+ from_: Literal["human", "gpt", "function_call", "observation"]
+ value: str
+
+ class Config:
+ fields: ClassVar[Dict[str, str]] = {"from_": "from"}
+ extra = "forbid"
+
+
+class ShareGPTData(BaseModel):
+ system: str
+ tools: str
+ conversations: List[ConversationItem]
+
+ class Config:
+ extra = "forbid"
+
+
+class ShareGPTDataCollector(BaseDataCollector):
+ def __init__(self) -> None:
+ super().__init__()
+ self.system_message: Optional[BaseMessage] = None
+ self.agent_name: Optional[str] = None
+ self.tools: List[FunctionTool] = []
+
+ def record(
+ self,
+ agent: Union[List[ChatAgent], ChatAgent],
+ ) -> Self:
+ r"""Inject an agent into the data collector."""
+ if not self.agent_name:
+ _agent = agent if isinstance(agent, ChatAgent) else agent[0]
+ self.agent_name = _agent.role_name
+ self.system_message = _agent._system_message
+ self.tools += list(_agent.tool_dict.values())
+
+ super().record(agent)
+ return self
+
+ def convert(self) -> Dict[str, Any]:
+ r"""Convert the collected data into a dictionary."""
+ if self.agent_name is None:
+ raise ValueError("No agent injected")
+
+ history = self.get_agent_history(self.agent_name)
+ if not history:
+ raise ValueError("No data collected.")
+
+ data = dict(
+ system=self.system_message.content if self.system_message else "",
+ tools=json.dumps(
+ [t.get_openai_tool_schema()["function"] for t in self.tools]
+ ),
+ conversations=[],
+ )
+
+ conversations: List[Any] = []
+ for _data in history:
+ role, message = _data.role, _data
+
+ if role == "user":
+ conversations.append(
+ {"from": "human", "value": message.message}
+ )
+ elif role == "assistant":
+ if message.function_call:
+ conversations.append(
+ {
+ "from": "function_call",
+ "value": json.dumps(message.function_call),
+ }
+ )
+ else:
+ conversations.append(
+ {"from": "gpt", "value": message.message}
+ )
+ elif role == "function":
+ conversations.append(
+ {
+ "from": "observation",
+ "value": json.dumps(message.message), # type: ignore[attr-defined]
+ }
+ )
+ data["conversations"] = conversations
+
+ self.data.append(data)
+ return data
+
+ def llm_convert(
+ self,
+ converter: Optional[OpenAISchemaConverter] = None,
+ prompt: Optional[str] = None,
+ ) -> Dict[str, Any]:
+ r"""Convert collected data using an LLM schema converter.
+
+ Args:
+ converter (Optional[OpenAISchemaConverter], optional):
+ The converter to use. (default: :obj:`OpenAISchemaConverter`)
+ prompt (Optional[str], optional): Prompt to guide the conversion.
+ (default: :obj:`DEFAULT_CONVERTER_PROMPTS`)
+
+ Returns:
+ Dict[str, str]: The converted data.
+
+ Raises:
+ ValueError: If no agent is injected or data cannot be collected.
+ """
+ prompt = prompt or DEFAULT_CONVERTER_PROMPTS
+ converter = converter or OpenAISchemaConverter()
+
+ system = self.system_message.content if self.system_message else ""
+ context = [f"System: {system}\n"]
+
+ context.append(
+ "Tools: "
+ + json.dumps(
+ [t.get_openai_tool_schema()["function"] for t in self.tools]
+ )
+ )
+ for _data in self.get_agent_history(str(self.agent_name)):
+ role, message = _data.role, _data
+ prefix = (
+ f"{role}: " if role != "user" else "User: " + f"{_data.name}: "
+ )
+ if message.function_call:
+ context.append(prefix + json.dumps(message.function_call))
+
+ elif role == "function":
+ context.append(prefix + json.dumps(message.message)) # type: ignore[attr-defined]
+ else:
+ context.append(prefix + str(message.message))
+ return converter.convert(
+ "\n".join(context), ShareGPTData, prompt
+ ).model_dump()
+
+ @staticmethod
+ def to_sharegpt_conversation(data: Dict[str, Any]) -> ShareGPTConversation:
+ messages = [
+ ShareGPTMessage(from_="system", value=data["system"]) # type: ignore[call-arg]
+ ]
+ for item in data["conversations"]:
+ messages.append(
+ ShareGPTMessage( # type: ignore[call-arg]
+ from_=FROM_HASH[item["from"]],
+ value=item["value"],
+ )
+ )
+ return ShareGPTConversation(root=messages)
diff --git a/camel/messages/__init__.py b/camel/messages/__init__.py
index 28e0fefd8..3a1c560da 100644
--- a/camel/messages/__init__.py
+++ b/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',
]
diff --git a/camel/schemas/openai_converter.py b/camel/schemas/openai_converter.py
index cf0df72a5..e534b1e5e 100644
--- a/camel/schemas/openai_converter.py
+++ b/camel/schemas/openai_converter.py
@@ -28,8 +28,8 @@ from camel.utils import (
from .base import BaseConverter
DEFAULT_CONVERTER_PROMPTS = """
- Extract key entities and attributes from the provided text,
- and convert them into a structured JSON format.
+ Extract key entities and attributes from the user
+ provided text, and convert them into a structured JSON format.
"""
diff --git a/examples/data_collector/alpaca_collector.py b/examples/data_collector/alpaca_collector.py
new file mode 100644
index 000000000..cbaedad83
--- /dev/null
+++ b/examples/data_collector/alpaca_collector.py
@@ -0,0 +1,68 @@
+# ========= 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.agents.chat_agent import ChatAgent
+from camel.configs.openai_config import ChatGPTConfig
+from camel.data_collector import AlpacaDataCollector
+from camel.messages.base import BaseMessage
+from camel.models.model_factory import ModelFactory
+from camel.types.enums import ModelPlatformType, ModelType
+
+model_config_dict = ChatGPTConfig(
+ temperature=0.0,
+).as_dict()
+
+model = ModelFactory.create(
+ model_platform=ModelPlatformType.DEFAULT,
+ model_type=ModelType.DEFAULT,
+ model_config_dict=model_config_dict,
+)
+
+agent = ChatAgent(
+ system_message=BaseMessage.make_assistant_message(
+ role_name="Tools calling operator",
+ content="You are a helpful assistant",
+ ),
+ model=model,
+)
+
+usr_msg = BaseMessage.make_user_message(
+ role_name="User",
+ content="When is the release date of the video game Portal?",
+)
+
+collector = AlpacaDataCollector().record(agent).start()
+
+# Automatically record the message
+resp = agent.step(usr_msg)
+
+print(collector.convert())
+
+print(collector.llm_convert())
+
+collector.reset()
+
+# Manually record the message
+collector.step("user", "Tools calling operator", usr_msg.content)
+
+collector.step("assistant", "Tools calling operator", resp.msgs[0].content)
+
+print(collector.convert())
+
+# ruff: noqa: E501
+"""
+{'instruction': 'You are a helpful assistant', 'input': 'When is the release date of the video game Portal?', 'output': 'The video game "Portal" was released on October 10, 2007, as part of the game bundle "The Orange Box," which also included "Half-Life 2" and its episodes. It was later released as a standalone game on April 9, 2008, for PC and Xbox 360.'}
+{'instruction': 'You are a helpful assistant', 'input': 'When is the release date of the video game Portal?', 'output': "The video game Portal was released on October 10, 2007, as part of the game bundle 'The Orange Box.' It was later released as a standalone game on April 9, 2008, for PC and Xbox 360."}
+{'instruction': 'You are a helpful assistant', 'input': 'When is the release date of the video game Portal?', 'output': 'The video game "Portal" was released on October 10, 2007, as part of the game bundle "The Orange Box," which also included "Half-Life 2" and its episodes. It was later released as a standalone game on April 9, 2008, for PC and Xbox 360.'}
+"""
diff --git a/examples/data_collector/sharegpt_collector.py b/examples/data_collector/sharegpt_collector.py
new file mode 100644
index 000000000..17c72227c
--- /dev/null
+++ b/examples/data_collector/sharegpt_collector.py
@@ -0,0 +1,105 @@
+# ========= 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 json
+
+from camel.agents import ChatAgent
+from camel.configs import ChatGPTConfig
+from camel.data_collector import ShareGPTDataCollector
+from camel.messages import BaseMessage
+from camel.models import ModelFactory
+from camel.toolkits import MathToolkit
+from camel.types import ModelPlatformType, ModelType
+
+tool_list = MathToolkit().get_tools()
+
+model_config_dict = ChatGPTConfig(
+ tools=tool_list,
+ temperature=0.0,
+).as_dict()
+
+model = ModelFactory.create(
+ model_platform=ModelPlatformType.DEFAULT,
+ model_type=ModelType.DEFAULT,
+ model_config_dict=model_config_dict,
+)
+
+# Set external_tools
+agent = ChatAgent(
+ system_message=BaseMessage.make_assistant_message(
+ role_name="Tools calling operator",
+ content="You are a helpful assistant",
+ ),
+ model=model,
+ tools=tool_list,
+)
+collector = ShareGPTDataCollector().record(agent).start()
+
+usr_msg = BaseMessage.make_user_message(
+ role_name="User",
+ content="Call tools to calculate 17 * 19 = ?",
+)
+
+# This will directly run the internal tool
+response = agent.step(usr_msg)
+
+
+print(json.dumps(collector.convert(), indent=4))
+print(json.dumps(collector.llm_convert(), indent=4))
+print(collector.to_sharegpt_conversation(collector.convert()))
+
+# ruff: noqa: E501
+"""
+{
+ "system": "You are a helpful assistant",
+ "tools": "[{\"name\": \"add\", \"description\": \"Adds two numbers.\", \"parameters\": {\"properties\": {\"a\": {\"type\": \"integer\", \"description\": \"The first number to be added.\"}, \"b\": {\"type\": \"integer\", \"description\": \"The second number to be added.\"}}, \"required\": [\"a\", \"b\"], \"type\": \"object\"}}, {\"name\": \"sub\", \"description\": \"Do subtraction between two numbers.\", \"parameters\": {\"properties\": {\"a\": {\"type\": \"integer\", \"description\": \"The minuend in subtraction.\"}, \"b\": {\"type\": \"integer\", \"description\": \"The subtrahend in subtraction.\"}}, \"required\": [\"a\", \"b\"], \"type\": \"object\"}}, {\"name\": \"mul\", \"description\": \"Multiplies two integers.\", \"parameters\": {\"properties\": {\"a\": {\"type\": \"integer\", \"description\": \"The multiplier in the multiplication.\"}, \"b\": {\"type\": \"integer\", \"description\": \"The multiplicand in the multiplication.\"}}, \"required\": [\"a\", \"b\"], \"type\": \"object\"}}]",
+ "conversations": [
+ {
+ "from": "human",
+ "value": "Call tools to calculate 17 * 19 = ?"
+ },
+ {
+ "from": "function_call",
+ "value": "{\"name\": \"mul\", \"arguments\": {\"a\": 17, \"b\": 19}}"
+ },
+ {
+ "from": "observation",
+ "value": "323"
+ },
+ {
+ "from": "gpt",
+ "value": "The result of \\( 17 \\times 19 \\) is \\( 323 \\)."
+ }
+ ]
+}
+{
+ "system": "You are a helpful assistant",
+ "tools": "[{\"name\": \"add\", \"description\": \"Adds two numbers.\", \"parameters\": {\"properties\": {\"a\": {\"type\": \"integer\", \"description\": \"The first number to be added.\"}, \"b\": {\"type\": \"integer\", \"description\": \"The second number to be added.\"}}, \"required\": [\"a\", \"b\"], \"type\": \"object\"}}, {\"name\": \"sub\", \"description\": \"Do subtraction between two numbers.\", \"parameters\": {\"properties\": {\"a\": {\"type\": \"integer\", \"description\": \"The minuend in subtraction.\"}, \"b\": {\"type\": \"integer\", \"description\": \"The subtrahend in subtraction.\"}}, \"required\": [\"a\", \"b\"], \"type\": \"object\"}}, {\"name\": \"mul\", \"description\": \"Multiplies two integers.\", \"parameters\": {\"properties\": {\"a\": {\"type\": \"integer\", \"description\": \"The multiplier in the multiplication.\"}, \"b\": {\"type\": \"integer\", \"description\": \"The multiplicand in the multiplication.\"}}, \"required\": [\"a\", \"b\"], \"type\": \"object\"}}]",
+ "conversations": [
+ {
+ "from_": "human",
+ "value": "Tools calling operator: Call tools to calculate 17 * 19 = ?"
+ },
+ {
+ "from_": "function_call",
+ "value": "{\"name\": \"mul\", \"arguments\": {\"a\": 17, \"b\": 19}}"
+ },
+ {
+ "from_": "gpt",
+ "value": "The result of \\( 17 \\times 19 \\) is \\( 323 \\)."
+ }
+ ]
+}
+root=[ShareGPTMessage(from_='system', value='You are a helpful assistant'), ShareGPTMessage(from_='human', value='Call tools to calculate 17 * 19 = ?'), ShareGPTMessage(from_='gpt', value='{"name": "multiply", "arguments": "{\'a\': 17, \'b\': 19}"}'), ShareGPTMessage(from_='human', value='"{\'result\': {\'323\'}}"'), ShareGPTMessage(from_='gpt', value='The result of \\( 17 \\times 19 \\) is \\( 323 \\).')]
+"""
diff --git a/test/data_collectors/test_alpaca_collector.py b/test/data_collectors/test_alpaca_collector.py
new file mode 100644
index 000000000..c7ab1bdca
--- /dev/null
+++ b/test/data_collectors/test_alpaca_collector.py
@@ -0,0 +1,96 @@
+# ========= 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.agents.chat_agent import ChatAgent
+from camel.configs.openai_config import ChatGPTConfig
+from camel.data_collector import AlpacaDataCollector
+from camel.messages.base import BaseMessage
+from camel.models.model_factory import ModelFactory
+from camel.types.enums import ModelPlatformType, ModelType
+
+
+def test_alpaca_converter():
+ model_config_dict = ChatGPTConfig(
+ temperature=0.0,
+ ).as_dict()
+
+ model = ModelFactory.create(
+ model_platform=ModelPlatformType.DEFAULT,
+ model_type=ModelType.DEFAULT,
+ model_config_dict=model_config_dict,
+ )
+
+ agent = ChatAgent(
+ system_message=BaseMessage.make_assistant_message(
+ role_name="Tools calling operator",
+ content="You are a helpful assistant. ",
+ ),
+ model=model,
+ )
+
+ usr_msg = BaseMessage.make_user_message(
+ role_name="User",
+ content="When is the release date of the video game Portal?",
+ )
+
+ collector = AlpacaDataCollector().record(agent).start()
+
+ _resp = agent.step(usr_msg)
+ resp = collector.convert()
+
+ assert (
+ resp["instruction"]
+ == "You are a helpful assistant. "
+ + "When is the release date of the video game Portal?"
+ )
+ assert resp['input'] == ''
+ assert resp['output'] != ''
+
+
+def test_alpaca_llm_converter():
+ model_config_dict = ChatGPTConfig(
+ temperature=0.0,
+ ).as_dict()
+
+ model = ModelFactory.create(
+ model_platform=ModelPlatformType.DEFAULT,
+ model_type=ModelType.DEFAULT,
+ model_config_dict=model_config_dict,
+ )
+
+ agent = ChatAgent(
+ system_message=BaseMessage.make_assistant_message(
+ role_name="Tools calling operator",
+ content="You are a helpful assistant. ",
+ ),
+ model=model,
+ )
+
+ usr_msg = BaseMessage.make_user_message(
+ role_name="User",
+ content="When is the release date of the video game Portal?",
+ )
+
+ collector = AlpacaDataCollector().record(agent).start()
+
+ _resp = agent.step(usr_msg)
+ resp = collector.llm_convert()
+
+ assert (
+ resp["instruction"]
+ == "You are a helpful assistant. "
+ + "When is the release date of the video game Portal?"
+ )
+ assert resp["input"] == ""
+ assert resp["output"] != ""
diff --git a/test/data_collectors/test_sharegpt_collector.py b/test/data_collectors/test_sharegpt_collector.py
new file mode 100644
index 000000000..3910cfcb6
--- /dev/null
+++ b/test/data_collectors/test_sharegpt_collector.py
@@ -0,0 +1,95 @@
+# ========= 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.agents import ChatAgent
+from camel.configs import ChatGPTConfig
+from camel.data_collector import ShareGPTDataCollector
+from camel.messages import BaseMessage
+from camel.models import ModelFactory
+from camel.toolkits import MathToolkit
+from camel.types import ModelPlatformType, ModelType
+
+
+def test_sharegpt_converter():
+ tool_list = MathToolkit().get_tools()
+
+ model_config_dict = ChatGPTConfig(
+ tools=tool_list,
+ temperature=0.0,
+ ).as_dict()
+
+ model = ModelFactory.create(
+ model_platform=ModelPlatformType.DEFAULT,
+ model_type=ModelType.DEFAULT,
+ model_config_dict=model_config_dict,
+ )
+
+ # Set external_tools
+ agent = ChatAgent(
+ system_message=BaseMessage.make_assistant_message(
+ role_name="Tools calling operator",
+ content="You are a helpful assistant",
+ ),
+ model=model,
+ tools=tool_list,
+ )
+ collector = ShareGPTDataCollector().record(agent).start()
+
+ usr_msg = BaseMessage.make_user_message(
+ role_name="User",
+ content="Call tools to calculate 17 * 19 = ?",
+ )
+
+ # This will directly run the internal tool
+ _ = agent.step(usr_msg)
+ resp = collector.convert()
+ assert resp["system"] == "You are a helpful assistant"
+ assert len(resp["conversations"]) == 4
+
+
+def test_sharegpt_llm_converter():
+ tool_list = MathToolkit().get_tools()
+
+ model_config_dict = ChatGPTConfig(
+ tools=tool_list,
+ temperature=0.0,
+ ).as_dict()
+
+ model = ModelFactory.create(
+ model_platform=ModelPlatformType.DEFAULT,
+ model_type=ModelType.DEFAULT,
+ model_config_dict=model_config_dict,
+ )
+
+ # Set external_tools
+ agent = ChatAgent(
+ system_message=BaseMessage.make_assistant_message(
+ role_name="Tools calling operator",
+ content="You are a helpful assistant",
+ ),
+ model=model,
+ tools=tool_list,
+ )
+ collector = ShareGPTDataCollector().record(agent).start()
+
+ usr_msg = BaseMessage.make_user_message(
+ role_name="User",
+ content="Call tools to calculate 17 * 19 = ?",
+ )
+
+ # This will directly run the internal tool
+ _ = agent.step(usr_msg)
+ resp = collector.llm_convert()
+ assert resp["system"] == "You are a helpful assistant"
+ assert len(resp["conversations"]) == 4