prompt-engineering-for-langgraph

Requirements

• langchain-core >= 0.3.0
• langchain-anthropic >= 1.1.0 (for ChatAnthropic with native structured output via json_schema method)
• langgraph >= 1.0.0 (for StateGraph examples)

Prompt Engineering for LangGraph

Effective prompts improve LLM behavior, accuracy, and reliability in LangGraph workflows.

ChatPromptTemplate (Recommended)

from langchain_core.prompts import ChatPromptTemplate
from langchain_anthropic import ChatAnthropic

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant specializing in {domain}."),
    ("human", "{user_input}")
])

# Initialize LLM
llm = ChatAnthropic(model="claude-sonnet-4-5-20250929")

# Use in chain
chain = prompt | llm
result = chain.invoke({"domain": "Python", "user_input": "Explain decorators"})

MessagesPlaceholder for Conversation Memory

MessagesPlaceholder enables dynamic injection of conversation history into prompts - critical for LangGraph stateful workflows.

from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder

# Basic usage with chat history
prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant."),
    MessagesPlaceholder(variable_name="chat_history", optional=True),
    ("human", "{input}")
])

# Invoke without history (optional=True allows this)
result = prompt.invoke({"input": "Hello!"})

# Invoke with history
result = prompt.invoke({
    "chat_history": [
        ("human", "What is 2+2?"),
        ("ai", "2+2 equals 4.")
    ],
    "input": "Now multiply that by 3"
})

MessagesPlaceholder Parameters

Parameter	Type	Default	Description
variable_name	str	required	Name of the variable for message list
optional	bool	False	If True, allows omitting the variable (returns empty list); if False, raises KeyError when variable missing
n_messages	int	None	Maximum number of messages to include