Skip to main content

Prompt Templates

A prompt template turns a dict of variables into a PromptValue the model can consume. ChatPromptTemplate is the one you'll use almost everywhere; PromptTemplate produces a plain string instead of a message list.

from langchain_core.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a concise technical writer."),
    ("human", "Explain {topic} in {sentences} sentences."),
])

prompt.invoke({"topic": "vector embeddings", "sentences": 2})

ChatPromptTemplate.from_messages accepts (role, template) tuples — "system", "human", "ai" — and fills each with the variables passed to invoke.

Conversation history: MessagesPlaceholder​

To splice a variable-length list of prior messages into a template (chat history, few-shot examples), use MessagesPlaceholder instead of a fixed slot:

from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant."),
    MessagesPlaceholder("history"),
    ("human", "{question}"),
])

prompt.invoke({
    "history": [("human", "My name is Ana"), ("ai", "Nice to meet you, Ana!")],
    "question": "What's my name?",
})

Partials and few-shot examples​

prompt.partial(...) binds a variable ahead of time, useful for values known at chain-build time (a system persona, a date) that shouldn't be re-passed on every invoke. For few-shot prompting, build the example block as a formatted string or a list fed through MessagesPlaceholder, so adding or swapping examples doesn't require rewriting the template.

Pitfalls

A raw { in a template — JSON inside a prompt, a code example with braces — is interpreted as a variable placeholder. Escape it as {{ and }}, or the template will raise a KeyError on the literal text.

See also​

  • Messages — the message objects a template ultimately produces.
  • Your First Chain — a template as the first stage of a chain.