langchain.agents.structured_chat.base.create_structured_chat_agent¶

langchain.agents.structured_chat.base.create_structured_chat_agent(llm: ~langchain_core.language_models.base.BaseLanguageModel, tools: ~typing.Sequence[~langchain_core.tools.BaseTool], prompt: ~langchain_core.prompts.chat.ChatPromptTemplate, tools_renderer: ~typing.Callable[[~typing.List[~langchain_core.tools.BaseTool]], str] = <function render_text_description_and_args>, *, stop_sequence: ~typing.Union[bool, ~typing.List[str]] = True) → Runnable[source]¶

Create an agent aimed at supporting tools with multiple inputs.

Parameters

  • llm (BaseLanguageModel) – LLM to use as the agent.

  • tools (Sequence[BaseTool]) – Tools this agent has access to.

  • prompt (ChatPromptTemplate) – The prompt to use. See Prompt section below for more.

  • stop_sequence (Union[bool, List[str]]) –

    bool or list of str. If True, adds a stop token of “Observation:” to avoid hallucinates. If False, does not add a stop token. If a list of str, uses the provided list as the stop tokens.

    Default is True. You may to set this to False if the LLM you are using does not support stop sequences.

  • tools_renderer (Callable[[List[BaseTool]], str]) – This controls how the tools are converted into a string and then passed into the LLM. Default is render_text_description.

Returns

A Runnable sequence representing an agent. It takes as input all the same input variables as the prompt passed in does. It returns as output either an AgentAction or AgentFinish.

Return type

Runnable

Examples

from langchain import hub
from langchain_community.chat_models import ChatOpenAI
from langchain.agents import AgentExecutor, create_structured_chat_agent

prompt = hub.pull("hwchase17/structured-chat-agent")
model = ChatOpenAI()
tools = ...

agent = create_structured_chat_agent(model, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools)

agent_executor.invoke({"input": "hi"})

# Using with chat history
from langchain_core.messages import AIMessage, HumanMessage
agent_executor.invoke(
    {
        "input": "what's my name?",
        "chat_history": [
            HumanMessage(content="hi! my name is bob"),
            AIMessage(content="Hello Bob! How can I assist you today?"),
        ],
    }
)

Prompt:

The prompt must have input keys:

  • tools: contains descriptions and arguments for each tool.

  • tool_names: contains all tool names.

  • agent_scratchpad: contains previous agent actions and tool outputs as a string.

Here’s an example:

from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder

system = '''Respond to the human as helpfully and accurately as possible. You have access to the following tools:

{tools}

Use a json blob to specify a tool by providing an action key (tool name) and an action_input key (tool input).

Valid "action" values: "Final Answer" or {tool_names}

Provide only ONE action per $JSON_BLOB, as shown:

```
{{
  "action": $TOOL_NAME,
  "action_input": $INPUT
}}
```

Follow this format:

Question: input question to answer
Thought: consider previous and subsequent steps
Action:
```
$JSON_BLOB
```
Observation: action result
... (repeat Thought/Action/Observation N times)
Thought: I know what to respond
Action:
```
{{
  "action": "Final Answer",
  "action_input": "Final response to human"
}}

Begin! Reminder to ALWAYS respond with a valid json blob of a single action. Use tools if necessary. Respond directly if appropriate. Format is Action:```$JSON_BLOB```then Observation'''

human = '''{input}

{agent_scratchpad}

(reminder to respond in a JSON blob no matter what)'''

prompt = ChatPromptTemplate.from_messages(
    [
        ("system", system),
        MessagesPlaceholder("chat_history", optional=True),
        ("human", human),
    ]
)

Examples using create_structured_chat_agent¶