PromptManager class

Package:

@microsoft/teams-ai

A filesystem based prompt manager.

Remarks

The default prompt manager uses the file system to define prompts that are compatible with Microsoft's Semantic Kernel SDK (see: https://github.com/microsoft/semantic-kernel)

Each prompt is a separate folder under a root prompts folder. The folder should contain the following files:

• "config.json": Required. Contains the prompts configuration and is a serialized instance of PromptTemplateConfig.
• "skprompt.txt": Required. Contains the text of the prompt and supports Semantic Kernels prompt template syntax.
• "actions.json": Optional. Contains a list of actions that can be called by the prompt.

Prompts can be loaded and used by name and new dynamically defined prompt templates can be registered with the prompt manager.

Constructors

PromptManager(PromptManagerOptions)

Creates a new 'PromptManager' instance.

Properties

options

Gets the configured prompt manager options.

Methods

addDataSource(DataSource)	Registers a new data source with the prompt manager.
addFunction(string, PromptFunction)	Registers a new prompt template function with the prompt manager.
addPrompt(PromptTemplate)	Registers a new prompt template with the prompt manager.
getDataSource(string)	Looks up a data source by name.
getFunction(string)	Looks up a prompt template function by name.
getPrompt(string)	Loads a named prompt template from the filesystem.
hasDataSource(string)	Checks for the existence of a named data source.
hasFunction(string)	Checks for the existence of a named prompt template function.
hasPrompt(string)	Checks for the existence of a named prompt.
invokeFunction(string, TurnContext, Memory, Tokenizer, string[])	Invokes a prompt template function by name.

Constructor Details

PromptManager(PromptManagerOptions)

Creates a new 'PromptManager' instance.

new PromptManager(options: PromptManagerOptions)

Parameters

options

PromptManagerOptions

Options used to configure the prompt manager.

Property Details

options

Gets the configured prompt manager options.

ConfiguredPromptManagerOptions options

Property Value

ConfiguredPromptManagerOptions

The configured prompt manager options.

Method Details

addDataSource(DataSource)

Registers a new data source with the prompt manager.

function addDataSource(dataSource: DataSource): PromptManager

Parameters

dataSource

DataSource

Data source to add.

Returns

PromptManager

The prompt manager for chaining.

addFunction(string, PromptFunction)

Registers a new prompt template function with the prompt manager.

function addFunction(name: string, fn: PromptFunction): PromptManager

Parameters

name

string

Name of the function to add.

fn

PromptFunction

Function to add.

Returns

PromptManager

• The prompt manager for chaining.

addPrompt(PromptTemplate)

Registers a new prompt template with the prompt manager.

function addPrompt(prompt: PromptTemplate): PromptManager

Parameters

prompt

PromptTemplate

Prompt template to add.

Returns

PromptManager

The prompt manager for chaining.

getDataSource(string)

Looks up a data source by name.

function getDataSource(name: string): DataSource

Parameters

name

string

Name of the data source to lookup.

Returns

DataSource

The data source.

getFunction(string)

Looks up a prompt template function by name.

function getFunction(name: string): PromptFunction

Parameters

name

string

Name of the function to lookup.

Returns

PromptFunction

The function.

getPrompt(string)

Loads a named prompt template from the filesystem.

function getPrompt(name: string): Promise<PromptTemplate>

Parameters

name

string

Name of the prompt to load.

Returns

Promise<PromptTemplate>

The loaded and parsed prompt template.

Remarks

The template will be pre-parsed and cached for use when the template is rendered by name.

Any augmentations will also be added to the template.

hasDataSource(string)

Checks for the existence of a named data source.

function hasDataSource(name: string): boolean

Parameters

name

string

Name of the data source to lookup.

Returns

boolean

True if the data source exists.

hasFunction(string)

Checks for the existence of a named prompt template function.

function hasFunction(name: string): boolean

Parameters

name

string

Name of the function to lookup.

Returns

boolean

True if the function exists.

hasPrompt(string)

Checks for the existence of a named prompt.

function hasPrompt(name: string): Promise<boolean>

Parameters

name

string

Name of the prompt to load.

Returns

Promise<boolean>

True if the prompt exists.

invokeFunction(string, TurnContext, Memory, Tokenizer, string[])

Invokes a prompt template function by name.

function invokeFunction(name: string, context: TurnContext, memory: Memory, tokenizer: Tokenizer, args: string[]): Promise<any>

Parameters

name

string

Name of the function to invoke.

context

TurnContext

Turn context for the current turn of conversation with the user.

memory

Memory

An interface for accessing state values.

tokenizer

Tokenizer

Tokenizer to use when rendering the prompt.

args

string[]

Arguments to pass to the function.

Returns

Promise<any>

Value returned by the function.