is there a way to set default tool when no tool is being selected?

[InsectScholar]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

the agent i am using
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain.agents import AgentType
from langchain.agents import AgentExecutor, create_openai_tools_agent

from langchain.memory import ConversationBufferMemory
from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
from langchain.chat_models import ChatOpenAI


tools=[retriever_tool2,retriever_tool]
prompt = ChatPromptTemplate.from_messages(
    [#You are a helpful NDIS chatbot, if you can't find answer in context say i dont know, never generate a wrong answer
       
        ("system", SystemPrompt),
        MessagesPlaceholder("chat_history", optional=True),
        ("human", "{input}"),
        MessagesPlaceholder("agent_scratchpad"),
    ])

memory = ConversationBufferMemory(
                            memory_key="chat_history",
                            max_len=4,
                            tools=tools,
                            return_messages=True,
                            k=1
                        )
#memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True, k=3)
agent=create_openai_tools_agent(llm,tools,prompt)
from langchain.agents import AgentExecutor
agent_executor=AgentExecutor(agent=agent,tools=tools,verbose=True,memory=memory)

Description

Langchain agent donot choose the tool every time I want to set a default tool for langchain agent such that even if it do not invoke any tool it should take that tool by default. Is there any way I can get it , by any way i mean any other agent, changes in original code or any thing i want it to be done
@DosuBot

System Info

Name: langchain-core
Version: 0.1.45

[dosubot[bot]]

Hey there, @InsectScholar! I'm here to help you with any bugs, questions, or contributions. Let's tackle this issue together!

Yes, there is a way to set a default tool for the LangChain agent such that even if it does not invoke any tool, it should use that tool by default. This can be achieved by using the ExceptionTool class, which is designed to handle cases where no specific tool is invoked.

Here is an example of how you can implement this:

from langchain.tools import BaseTool
from langchain.agents import AgentExecutor, create_openai_tools_agent
from langchain.memory import ConversationBufferMemory
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain.chat_models import ChatOpenAI
from typing import Optional

class ExceptionTool(BaseTool):
    """Tool that just returns the query."""
    name: str = "_Exception"
    description: str = "Exception tool"

    def _run(self, query: str, run_manager: Optional[CallbackManagerForToolRun] = None) -> str:
        return query

    async def _arun(self, query: str, run_manager: Optional[AsyncCallbackManagerForToolRun] = None) -> str:
        return query

# Define your tools
tools = [retriever_tool2, retriever_tool]

# Create the prompt
prompt = ChatPromptTemplate.from_messages([
    ("system", SystemPrompt),
    MessagesPlaceholder("chat_history", optional=True),
    ("human", "{input}"),
    MessagesPlaceholder("agent_scratchpad"),
])

# Create the memory
memory = ConversationBufferMemory(
    memory_key="chat_history",
    max_len=4,
    tools=tools,
    return_messages=True,
    k=1
)

# Create the agent
llm = ChatOpenAI()  # Assuming you have defined your LLM
agent = create_openai_tools_agent(llm, tools, prompt)

# Create the agent executor with the ExceptionTool as the default tool
agent_executor = AgentExecutor(agent=agent, tools=tools + [ExceptionTool()], verbose=True, memory=memory)

In this setup, the ExceptionTool will be used as the default tool when no other tool is invoked by the agent ^[1].

To continue talking to Dosu, mention @dosu.

[InsectScholar]

if i want to set retriever_tool2 as ExceptionTool is it possible? @dosu.

[dosubot[bot]]

Yes, you can set retriever_tool2 as the default tool by modifying the logic to use retriever_tool2 when no specific tool is invoked. Here’s how you can do it:

1. Modify the create_openai_tools_agent function to accept a default tool parameter.
2. Ensure the agent uses the default tool when no specific tool is invoked.

Here is an example implementation:

from langchain.tools import BaseTool
from langchain.agents import AgentExecutor, create_openai_tools_agent
from langchain.memory import ConversationBufferMemory
from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder
from langchain.chat_models import ChatOpenAI
from typing import Optional, Sequence, List, Any

# Define your tools
tools = [retriever_tool2, retriever_tool]

# Create the prompt
prompt = ChatPromptTemplate.from_messages([
    ("system", SystemPrompt),
    MessagesPlaceholder("chat_history", optional=True),
    ("human", "{input}"),
    MessagesPlaceholder("agent_scratchpad"),
])

# Create the memory
memory = ConversationBufferMemory(
    memory_key="chat_history",
    max_len=4,
    tools=tools,
    return_messages=True,
    k=1
)

# Create the agent with a default tool
def create_openai_tools_agent_with_default(
    llm: BaseLanguageModel,
    tools: Sequence[BaseTool],
    default_tool: BaseTool,
    callback_manager: Optional[BaseCallbackManager] = None,
    extra_prompt_messages: Optional[List[BaseMessagePromptTemplate]] = None,
    system_message: Optional[SystemMessage] = SystemMessage(
        content="You are a helpful AI assistant."
    ),
    **kwargs: Any,
) -> BaseMultiActionAgent:
    prompt = OpenAIMultiFunctionsAgent.create_prompt(
        extra_prompt_messages=extra_prompt_messages,
        system_message=system_message,
    )
    
    agent = OpenAIMultiFunctionsAgent(
        llm=llm,
        prompt=prompt,
        tools=tools,
        callback_manager=callback_manager,
        **kwargs,
    )
    
    # Set the default tool
    agent.default_tool = default_tool
    
    return agent

# Create the agent
llm = ChatOpenAI()  # Assuming you have defined your LLM
agent = create_openai_tools_agent_with_default(llm, tools, default_tool=retriever_tool2)

# Create the agent executor
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True, memory=memory)

In this setup, retriever_tool2 is set as the default tool, and it will be used when no specific tool is invoked by the agent ^[1][2].

To continue talking to Dosu, mention @dosu.

[InsectScholar]

base tool does not help can i change something in agent since agent returns the tool of choice @dosu.

[dosubot[bot]]

To modify your LangChain agent so that it defaults to retriever_tool2 when no specific tool is invoked, you can adjust the get_tools function to include retriever_tool2 as a default tool. Here is how you can do it:

1. Define retriever_tool2.
2. Modify the get_tools function to include retriever_tool2 by default.

Here is the updated code:

from langchain.tools.retriever import create_retriever_tool
from langchain_core.documents import Document

# Define retriever_tool2
retriever_tool2 = create_retriever_tool(
    retriever=ArxivRetriever(),  # Replace with the actual retriever for retriever_tool2
    name="RetrieverTool2",
    description="Useful for retrieving information when no specific tool is invoked"
)

# Update the get_tools function
def get_tools(query):
    docs = retriever.invoke(query)
    tools = [ALL_TOOLS[d.metadata["index"]] for d in docs]
    if not tools:
        tools.append(retriever_tool2)
    return tools

This ensures that retriever_tool2 is included in the list of tools if no other tools are retrieved based on the query.

Additionally, you can use the high-level create_openai_tools_agent API to construct the agent, ensuring that retriever_tool2 is always available as a fallback tool:

from langchain_openai import ChatOpenAI
from langchain.agents import AgentExecutor, create_openai_tools_agent

llm = ChatOpenAI(temperature=0)
tools = [retriever_tool2]  # Add other tools as needed

agent = create_openai_tools_agent(llm, tools)
agent_executor = AgentExecutor(agent=agent, tools=tools)

# Now you can invoke the agent
result = agent_executor.invoke({"input": "your query here"})
print(result["output"])

This setup ensures that retriever_tool2 is always included and used as a default when no specific tool is invoked ^[1][2][3].

To continue talking to Dosu, mention @dosu.

[Joong-Sunny]

very helpful

[oussamaJmaaa]

did you find a solution for this ?

[eyurtsev]

Depending on which model provider you're using you may be able to force the chat model to use a tool (e.g., https://community.openai.com/t/new-api-feature-forcing-function-calling-via-tool-choice-required/731488/2)

[oussamaJmaaa]

i'm using langchain ChatOpenai() ,

llm = ChatOpenAI(
    model_name='gpt-4o-mini',
    temperature=0.0,
)

tools = [
    Tool(
        name='task1',
        func=ge_task1,
        description=(
            "Useful for task1."
        ),
    ),
    Tool(
        name="task2",
        func=get_task2,
        description="Useful for task2."
    ),
    Tool(
        name="task3",
        func=get_task3,
        description="Useful for task3."
    ),
]

agent = initialize_agent(
    agent=AgentType.CONVERSATIONAL_REACT_DESCRIPTION,
    tools=tools,
    llm=llm,
    verbose=True,
    max_iterations=3,
    #early_stopping_method='generate',
    memory=conversational_memory
)

i couldn't find a way to include a default prompt template or enforce a default behavior to the agent . when it gets asked about its name it say im a model from openai and it also responds to general questions which is not practical .
if you have an idea on how to solve this please inform me .