Prompt Templates

Overview

Prompt templates allow you to create reusable prompts with variable substitution. LangChain provides templates for both string-based prompts and chat-based prompts.

BasePromptTemplate

Base class for all prompt templates. Source: langchain_core.prompts.base:39 Inherits: RunnableSerializable[dict, PromptValue]

Type Parameters

FormatOutputType

TypeVar

The output type when formatting the prompt

Properties

input_variables

list[str]

required

List of variable names required as inputs to the prompt

optional_variables

list[str]

default:"[]"

List of optional variable names (auto-inferred from placeholders)

input_types

dict[str, Any]

default:"{}"

Dictionary mapping variable names to their expected types. Defaults to str for all.

output_parser

BaseOutputParser | None

default:"None"

Optional output parser to apply to the LLM response

partial_variables

Mapping[str, Any]

default:"{}"

Partial variables that are pre-filled

metadata

dict[str, Any] | None

default:"None"

Metadata for tracing

Methods

invoke

def invoke(
    self,
    input: dict,
    config: RunnableConfig | None = None
) -> PromptValue

Format the prompt with input variables.

input

dict

required

Dictionary of variable values

return

PromptValue

Formatted prompt value

format_prompt

def format_prompt(self, **kwargs: Any) -> PromptValue

Format the prompt with keyword arguments.

partial

def partial(self, **kwargs: Any) -> BasePromptTemplate

Return a partial prompt template with some variables pre-filled.

**kwargs

Any

required

Variables to pre-fill

return

BasePromptTemplate

New prompt template with partial variables

PromptTemplate

String prompt template using Python’s str.format syntax. Source: langchain_core.prompts.prompt Inherits: BasePromptTemplate

Properties

template

str

required

The template string with {variable} placeholders

template_format

Literal['f-string', 'jinja2', 'mustache']

default:"'f-string'"

Format of the template string

validate_template

bool

default:"False"

Whether to validate the template on initialization

Constructor

@classmethod
def from_template(
    cls,
    template: str,
    *,
    template_format: str = "f-string",
    partial_variables: dict[str, Any] | None = None,
    **kwargs: Any
) -> PromptTemplate

Create a prompt template from a template string.

template

str

required

Template string with variables in {variable} format

template_format

str

default:"'f-string'"

Template format ('f-string', 'jinja2', or 'mustache')

partial_variables

dict | None

Variables to pre-fill

Example

from langchain_core.prompts import PromptTemplate

prompt = PromptTemplate.from_template(
    "Tell me a joke about {topic}"
)
prompt.invoke({"topic": "cats"})  # PromptValue
prompt.format(topic="dogs")  # "Tell me a joke about dogs"

ChatPromptTemplate

Template for chat-based prompts using messages. Source: langchain_core.prompts.chat Inherits: BasePromptTemplate

Properties

messages

list[MessageLike]

required

List of message templates or placeholders

Constructor

@classmethod
def from_messages(
    cls,
    messages: Sequence[MessageLike | tuple[str, str]]
) -> ChatPromptTemplate

Create a chat prompt template from a list of messages.

messages

Sequence

required

List of message templates. Each can be:

A tuple like ("human", "Hello {name}")
A BaseMessage instance
A BaseMessagePromptTemplate
A MessagesPlaceholder

Example

from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder

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

result = prompt.invoke({
    "history": [("human", "Hi"), ("ai", "Hello!")],
    "input": "What's the weather?"
})

Class Methods

from_template

@classmethod
def from_template(
    cls,
    template: str,
    *,
    role: str = "human",
    **kwargs: Any
) -> ChatPromptTemplate

Create a chat prompt with a single message.

template

str

required

Template string

role

str

default:"'human'"

Role of the message ('human', 'ai', 'system', etc.)

MessagesPlaceholder

Placeholder for inserting a list of messages. Source: langchain_core.prompts.chat:52 Inherits: BaseMessagePromptTemplate

Properties

variable_name

str

required

Name of the variable containing messages

optional

bool

default:"False"

Whether the placeholder is optional. If True, returns empty list when variable not provided.

n_messages

PositiveInt | None

default:"None"

Maximum number of messages to include. If None, includes all.

Constructor

def __init__(
    self,
    variable_name: str,
    *,
    optional: bool = False,
    **kwargs: Any
)

variable_name

str

required

Name of variable containing messages

optional

bool

default:"False"

Whether the variable is optional

Example

from langchain_core.prompts import ChatPromptTemplate, MessagesPlaceholder

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

prompt.invoke({
    "chat_history": [
        ("human", "What's 2+2?"),
        ("ai", "4")
    ],
    "question": "What's 3+3?"
})

HumanMessagePromptTemplate

Template for creating human messages. Source: langchain_core.prompts.message Inherits: BaseMessagePromptTemplate

Properties

prompt

StringPromptTemplate

required

The underlying string prompt template

Constructor

@classmethod
def from_template(
    cls,
    template: str,
    *,
    template_format: str = "f-string",
    **kwargs: Any
) -> HumanMessagePromptTemplate

AIMessagePromptTemplate

Template for creating AI messages. Source: langchain_core.prompts.message Inherits: BaseMessagePromptTemplate

SystemMessagePromptTemplate

Template for creating system messages. Source: langchain_core.prompts.message Inherits: BaseMessagePromptTemplate

FewShotPromptTemplate

Template for few-shot learning with examples. Source: langchain_core.prompts.few_shot Inherits: BasePromptTemplate

Properties

examples

list[dict] | None

List of example dictionaries

example_selector

BaseExampleSelector | None

Selector to choose examples dynamically

example_prompt

PromptTemplate

required

Template for formatting each example

suffix

str

required

Suffix template appended after examples

prefix

str

default:"''"

Prefix template prepended before examples

example_separator

str

default:"'\\\\n\\\\n'"

String separator between examples

Example

from langchain_core.prompts import FewShotPromptTemplate, PromptTemplate

example_prompt = PromptTemplate.from_template(
    "Question: {question}\nAnswer: {answer}"
)

prompt = FewShotPromptTemplate(
    examples=[
        {"question": "What's 2+2?", "answer": "4"},
        {"question": "What's 3+3?", "answer": "6"}
    ],
    example_prompt=example_prompt,
    suffix="Question: {input}\nAnswer:",
    input_variables=["input"]
)

prompt.format(input="What's 4+4?")

PipelinePromptTemplate

Compose multiple prompt templates in a pipeline. Source: langchain_core.prompts.pipeline Inherits: BasePromptTemplate

Properties

final_prompt

BasePromptTemplate

required

Final prompt template to format

pipeline_prompts

list[tuple[str, BasePromptTemplate]]

required

List of (name, prompt) tuples to execute in order

ImagePromptTemplate

Template for creating prompts with images. Source: langchain_core.prompts.image Inherits: BasePromptTemplate

Properties

template

str | dict

required

Image URL template or image data template

Example

from langchain_core.prompts import ChatPromptTemplate
from langchain_core.prompts.image import ImagePromptTemplate

prompt = ChatPromptTemplate.from_messages([
    ("human", [
        {"type": "text", "text": "What's in this image?"},
        {"type": "image", "source": {"type": "url", "url": "{image_url}"}}
    ])
])

prompt.invoke({"image_url": "https://example.com/image.jpg"})

Core

LangChain

Overview

BasePromptTemplate

Type Parameters

Properties

Methods

invoke

format_prompt

partial

PromptTemplate

Properties

Constructor

Example

ChatPromptTemplate

Properties

Constructor

Example

Class Methods

from_template

MessagesPlaceholder

Properties

Constructor

Example

HumanMessagePromptTemplate

Properties

Constructor

AIMessagePromptTemplate

SystemMessagePromptTemplate

FewShotPromptTemplate

Properties

Example

PipelinePromptTemplate

Properties

ImagePromptTemplate

Properties

Example

Build docs developers (and LLMs) love

Core

LangChain

​Overview

​BasePromptTemplate

​Type Parameters

​Properties

​Methods

​invoke

​format_prompt

​partial

​PromptTemplate

​Properties

​Constructor

​Example

​ChatPromptTemplate

​Properties

​Constructor

​Example

​Class Methods

​from_template

​MessagesPlaceholder

​Properties

​Constructor

​Example

​HumanMessagePromptTemplate

​Properties

​Constructor

​AIMessagePromptTemplate

​SystemMessagePromptTemplate

​FewShotPromptTemplate

​Properties

​Example

​PipelinePromptTemplate

​Properties

​ImagePromptTemplate

​Properties

​Example

Build docs developers (and LLMs) love

Overview

BasePromptTemplate

Type Parameters

Properties

Methods

invoke

format_prompt

partial

PromptTemplate

Properties

Constructor

Example

ChatPromptTemplate

Properties

Constructor

Example

Class Methods

from_template

MessagesPlaceholder

Properties

Constructor

Example

HumanMessagePromptTemplate

Properties

Constructor

AIMessagePromptTemplate

SystemMessagePromptTemplate

FewShotPromptTemplate

Properties

Example

PipelinePromptTemplate

Properties

ImagePromptTemplate

Properties

Example