Implementation:Langchain ai Langchain Conversation Loop Pattern

Knowledge Sources	LangChain
Domains	Agentic_AI, Conversation_Management
Last Updated	2026-02-11 00:00 GMT

Overview

Pattern for implementing the agent conversation loop using BaseChatModel.invoke() with full message history provided by langchain-core.

Description

The conversation loop pattern uses BaseChatModel.invoke() with a growing list of messages. After each tool execution round, the AIMessage (with tool_calls) and ToolMessages (with results) are appended to the message history before re-invocation. The same invoke() path handles both tool-calling and final-answer responses.

Usage

This is a user-defined pattern (not a single API call). Implement it manually or use LangGraph for managed agent loops.

Code Reference

Source Location

• Repository: langchain
• File: libs/core/langchain_core/language_models/chat_models.py
• Lines: L389-437 (invoke — same path re-invoked with history)

Signature

# Re-invocation uses the same invoke() signature:
def invoke(
    self,
    input: LanguageModelInput,
    config: RunnableConfig | None = None,
    *,
    stop: list[str] | None = None,
    **kwargs: Any,
) -> AIMessage:

Import

from langchain_core.messages import HumanMessage, AIMessage, ToolMessage

I/O Contract

Inputs

Name	Type	Required	Description
input	list[BaseMessage]	Yes	Full conversation: [HumanMessage, AIMessage(tool_calls), ToolMessage, ...]

Outputs

Name	Type	Description
return	AIMessage	Either another round of tool_calls or a final text response

Usage Examples

Complete Agent Loop

from langchain_openai import ChatOpenAI
from langchain_core.messages import HumanMessage, ToolMessage
from langchain_core.tools import tool

@tool
def get_weather(city: str) -> str:
    """Get the current weather for a city."""
    return f"72°F and sunny in {city}"

llm = ChatOpenAI(model="gpt-4o-mini").bind_tools([get_weather])

messages = [HumanMessage(content="What's the weather in SF and NYC?")]

# Agent loop
for _ in range(5):  # Max iterations
    response = llm.invoke(messages)
    messages.append(response)

    if not response.tool_calls:
        print("Final answer:", response.content)
        break

    # Execute tool calls
    for tc in response.tool_calls:
        result = get_weather.invoke(tc)
        messages.append(ToolMessage(
            content=str(result),
            tool_call_id=tc["id"],
            name=tc["name"],
        ))

Related Pages

Implements Principle

• Principle:Langchain_ai_Langchain_Conversation_Looping