Messages

A chat model's input and output are both lists of typed message objects, not raw strings. Each message type maps to a role.

Type	Role	Written by	Purpose
`SystemMessage`	`system`	you, once per chain/conversation	instructions and persona, not user-visible
`HumanMessage`	`user`	your application, from user input	the user's turn
`AIMessage`	`assistant`	the model	the model's reply — may carry `tool_calls` instead of (or alongside) text
`ToolMessage`	`tool`	your application, after executing a tool	the result of a tool call, keyed to a `tool_call_id`

from langchain.messages import AIMessage, HumanMessage, SystemMessage, ToolMessage

messages = [
    SystemMessage("You can call the get_weather tool."),
    HumanMessage("What's the weather in Lisbon?"),
    AIMessage(
        content="",
        tool_calls=[{"name": "get_weather", "args": {"city": "Lisbon"}, "id": "call_1"}],
    ),
    ToolMessage(content="18°C, partly cloudy", tool_call_id="call_1"),
]

response = model.invoke(messages)  # model now answers using the tool result

An AIMessage carrying tool_calls is the shape the tool-calling loop depends on — see Tool Calling.

See also​

See also