Споделяне чрез

Create your own prompts to gather user input

  • Статия

APPLIES TO: SDK v4

A conversation between a bot and a user often involves asking (prompting) the user for information, parsing the user's response, and then acting on that information. Your bot should track the context of a conversation, so that it can manage its behavior and remember answers to previous questions. A bot's state is information it tracks to respond appropriately to incoming messages.

Tip

The dialogs library provides built-in prompts that provide more functionality that users can use. Examples of those prompts can be found in the Implement sequential conversation flow article.

Note

The Bot Framework JavaScript, C#, and Python SDKs will continue to be supported, however, the Java SDK is being retired with final long-term support ending in November 2023.

Existing bots built with the Java SDK will continue to function.

For new bot building, consider using Microsoft Copilot Studio and read about choosing the right copilot solution.

For more information, see The future of bot building.

Prerequisites

About the sample code

The sample bot asks the user a series of questions, validates some of their answers, and saves their input. The following diagram shows the relationship between the bot, user profile, and conversation flow classes.

Class diagram for the C# sample.

  • A UserProfile class for the user information that the bot will collect.
  • A ConversationFlow class to control our conversation state while gathering user information.
  • An inner ConversationFlow.Question enumeration for tracking where you are in the conversation.

The user state will track the user's name, age, and chosen date, and conversation state will track what you last asked the user. Since you don't plan to deploy this bot, you'll configure user and conversation state to use memory storage.

You use the bot's message turn handler plus user and conversation state properties to manage the flow of the conversation and the collection of input. In your bot, you'll record the state property information received during each iteration of the message turn handler.

Create conversation and user objects

Create the user and conversation state objects at startup and consume them via dependency injection in the bot constructor.

Startup.cs

// Create the Bot Adapter with error handling enabled.
services.AddSingleton<IBotFrameworkHttpAdapter, AdapterWithErrorHandler>();

// Create the storage we'll be using for User and Conversation state. (Memory is great for testing purposes.)
services.AddSingleton<IStorage, MemoryStorage>();

// Create the User state.
services.AddSingleton<UserState>();

// Create the Conversation state.
services.AddSingleton<ConversationState>();

Bots/CustomPromptBot.cs

private readonly BotState _userState;
private readonly BotState _conversationState;

public CustomPromptBot(ConversationState conversationState, UserState userState)
{
    _conversationState = conversationState;
    _userState = userState;
}

Create property accessors

Create property accessors for the user profile and conversation flow properties and then call GetAsync to retrieve the property value from state.

Bots/CustomPromptBot.cs

protected override async Task OnMessageActivityAsync(ITurnContext<IMessageActivity> turnContext, CancellationToken cancellationToken)
{
    var conversationStateAccessors = _conversationState.CreateProperty<ConversationFlow>(nameof(ConversationFlow));
    var flow = await conversationStateAccessors.GetAsync(turnContext, () => new ConversationFlow(), cancellationToken);

    var userStateAccessors = _userState.CreateProperty<UserProfile>(nameof(UserProfile));
    var profile = await userStateAccessors.GetAsync(turnContext, () => new UserProfile(), cancellationToken);

Before the turn ends, call SaveChangesAsync to write any state changes to storage.

    await _conversationState.SaveChangesAsync(turnContext, false, cancellationToken);
    await _userState.SaveChangesAsync(turnContext, false, cancellationToken);
}

Message turn handler

When handling message activities, the message handler uses a helper method to manage the conversation and prompt the user. The helper method is described in the following section.

Bots/CustomPromptBot.cs

protected override async Task OnMessageActivityAsync(ITurnContext<IMessageActivity> turnContext, CancellationToken cancellationToken)
{
    var conversationStateAccessors = _conversationState.CreateProperty<ConversationFlow>(nameof(ConversationFlow));
    var flow = await conversationStateAccessors.GetAsync(turnContext, () => new ConversationFlow(), cancellationToken);

    var userStateAccessors = _userState.CreateProperty<UserProfile>(nameof(UserProfile));
    var profile = await userStateAccessors.GetAsync(turnContext, () => new UserProfile(), cancellationToken);

    await FillOutUserProfileAsync(flow, profile, turnContext, cancellationToken);

    // Save changes.
    await _conversationState.SaveChangesAsync(turnContext, false, cancellationToken);
    await _userState.SaveChangesAsync(turnContext, false, cancellationToken);
}

Filling out the user profile

The bot prompts the user for information, based on which question, if any, that the bot asked on the previous turn. Input is parsed using a validation method.

Each validation method follows a similar design:

  • The return value indicates whether the input is a valid answer for this question.
  • If validation passes, it produces a parsed and normalized value to save.
  • If validation fails, it produces a message with which the bot can ask for the information again.

The validation methods are described in the following section.

Bots/CustomPromptBot.cs

{
    var input = turnContext.Activity.Text?.Trim();
    string message;

    switch (flow.LastQuestionAsked)
    {
        case ConversationFlow.Question.None:
            await turnContext.SendActivityAsync("Let's get started. What is your name?", null, null, cancellationToken);
            flow.LastQuestionAsked = ConversationFlow.Question.Name;
            break;
        case ConversationFlow.Question.Name:
            if (ValidateName(input, out var name, out message))
            {
                profile.Name = name;
                await turnContext.SendActivityAsync($"Hi {profile.Name}.", null, null, cancellationToken);
                await turnContext.SendActivityAsync("How old are you?", null, null, cancellationToken);
                flow.LastQuestionAsked = ConversationFlow.Question.Age;
                break;
            }
            else
            {
                await turnContext.SendActivityAsync(message ?? "I'm sorry, I didn't understand that.", null, null, cancellationToken);
                break;
            }

        case ConversationFlow.Question.Age:
            if (ValidateAge(input, out var age, out message))
            {
                profile.Age = age;
                await turnContext.SendActivityAsync($"I have your age as {profile.Age}.", null, null, cancellationToken);
                await turnContext.SendActivityAsync("When is your flight?", null, null, cancellationToken);
                flow.LastQuestionAsked = ConversationFlow.Question.Date;
                break;
            }
            else
            {
                await turnContext.SendActivityAsync(message ?? "I'm sorry, I didn't understand that.", null, null, cancellationToken);
                break;
            }

        case ConversationFlow.Question.Date:
            if (ValidateDate(input, out var date, out message))
            {
                profile.Date = date;
                await turnContext.SendActivityAsync($"Your cab ride to the airport is scheduled for {profile.Date}.");
                await turnContext.SendActivityAsync($"Thanks for completing the booking {profile.Name}.");
                await turnContext.SendActivityAsync($"Type anything to run the bot again.");
                flow.LastQuestionAsked = ConversationFlow.Question.None;
                profile = new UserProfile();
                break;
            }
            else
            {
                await turnContext.SendActivityAsync(message ?? "I'm sorry, I didn't understand that.", null, null, cancellationToken);
                break;
            }
    }
}

Parse and validate input

The bot uses the following criteria to validate input.

  • The name must be a non-empty string. It's normalized by trimming white-space.
  • The age must be between 18 and 120. It's normalized by returning an integer.
  • The date must be any date or time at least an hour in the future. It's normalized by returning just the date portion of the parsed input.

Note

For the age and date input, the sample uses the Microsoft/Recognizers-Text libraries to perform the initial parsing. This is just one way to parse the input. For more information about these libraries, see the project's README.

Bots/CustomPromptBot.cs

private static bool ValidateName(string input, out string name, out string message)
{
    name = null;
    message = null;

    if (string.IsNullOrWhiteSpace(input))
    {
        message = "Please enter a name that contains at least one character.";
    }
    else
    {
        name = input.Trim();
    }

    return message is null;
}

private static bool ValidateAge(string input, out int age, out string message)
{
    age = 0;
    message = null;

    // Try to recognize the input as a number. This works for responses such as "twelve" as well as "12".
    try
    {
        // Attempt to convert the Recognizer result to an integer. This works for "a dozen", "twelve", "12", and so on.
        // The recognizer returns a list of potential recognition results, if any.

        var results = NumberRecognizer.RecognizeNumber(input, Culture.English);

        foreach (var result in results)
        {
            // The result resolution is a dictionary, where the "value" entry contains the processed string.
            if (result.Resolution.TryGetValue("value", out var value))
            {
                age = Convert.ToInt32(value);
                if (age >= 18 && age <= 120)
                {
                    return true;
                }
            }
        }

        message = "Please enter an age between 18 and 120.";
    }
    catch
    {
        message = "I'm sorry, I could not interpret that as an age. Please enter an age between 18 and 120.";
    }

    return message is null;
}

private static bool ValidateDate(string input, out string date, out string message)
{
    date = null;
    message = null;

    // Try to recognize the input as a date-time. This works for responses such as "11/14/2018", "9pm", "tomorrow", "Sunday at 5pm", and so on.
    // The recognizer returns a list of potential recognition results, if any.
    try
    {
        var results = DateTimeRecognizer.RecognizeDateTime(input, Culture.English);

        // Check whether any of the recognized date-times are appropriate,
        // and if so, return the first appropriate date-time. We're checking for a value at least an hour in the future.
        var earliest = DateTime.Now.AddHours(1.0);

        foreach (var result in results)
        {
            // The result resolution is a dictionary, where the "values" entry contains the processed input.
            var resolutions = result.Resolution["values"] as List<Dictionary<string, string>>;

            foreach (var resolution in resolutions)
            {
                // The processed input contains a "value" entry if it is a date-time value, or "start" and
                // "end" entries if it is a date-time range.
                if (resolution.TryGetValue("value", out var dateString)
                    || resolution.TryGetValue("start", out dateString))
                {
                    if (DateTime.TryParse(dateString, out var candidate)
                        && earliest < candidate)
                    {
                        date = candidate.ToShortDateString();
                        return true;
                    }
                }
            }
        }

        message = "I'm sorry, please enter a date at least an hour out.";
    }
    catch
    {
        message = "I'm sorry, I could not interpret that as an appropriate date. Please enter a date at least an hour out.";
    }

    return false;
}

Test the bot locally

Download and install the Bot Framework Emulator to test the bot locally.

  1. Run the sample locally on your machine. If you need instructions, refer to the README file for C# sample, JS sample, or the Python sample.
  2. Test it using the Emulator.

Additional resources

The Dialogs library provides classes that automate many aspects of managing conversations.

Next step

Implement sequential conversation flow