Botframework 聊天机器人中的空响应和跨会话问题
我正在使用V4的bot框架。使用API调用，我试图获取数据并将其显示给用户，我定义了一个自定义方法从API中捕获数据并对其进行预处理，然后通过瀑布式对话框将其发送给用户，我使该方法异步，并使用wait where调用它 我面临的问题有两种情况-Botframework 聊天机器人中的空响应和跨会话问题,botframework,Botframework,我正在使用V4的bot框架。使用API调用，我试图获取数据并将其显示给用户，我定义了一个自定义方法从API中捕获数据并对其进行预处理，然后通过瀑布式对话框将其发送给用户，我使该方法异步，并使用wait where调用它 我面临的问题有两种情况- 当两个用户向一个实例发送问题时，其中一个响应捕获为null，而不是从API获取的值 我们使用提示卡来显示结果，当用户点击按钮时，会偶尔观察到交叉会话。 非常感谢您在这方面的帮助 定义用于进行API调用并获取数据的自定义方法： 公共静态异步任务 Ge
- 当两个用户向一个实例发送问题时，其中一个响应捕获为null，而不是从API获取的值
- 我们使用提示卡来显示结果，当用户点击按钮时，会偶尔观察到交叉会话。 非常感谢您在这方面的帮助
公共静态异步任务
GetEPCallsDoneSync（ConversationData ConversationData）
{
LogWriter.LogWrite（“信息：访问端点”）；
字符串responseMessage=null；
尝试
{
conversationData.fulFillmentMap=等待
AnchorUtil.GetFulfillmentAsync（xxxxx）；//从API调用获取数据的响应
if（conversationData.fulFillmentMap==null | | |（conversationData.fulFillmentMap.ContainsKey（“状态”）&&conversationData.fulFillmentMap[“状态”].ToString（）！=“200”））
{
responseMessage=“抱歉，出现问题。请稍后再试！”；
}
其他的
{
conversationData.NLGresultMap=wait
AnchorUtil.GetNLGAsync（conversationData.fulFillmentMap，xxxx）；//API调用以获取要显示的响应
if（conversationData.errorCaptureDict.ContainsKey（“实现中间错误”）| | conversationData.NLGresultMap.ContainsKey（“NLGError”））
{
responseMessage=“抱歉，出现问题：（请稍后再试！！！”；
}
其他的
{
responseMessage=FormatDataResponse（conversationData.NLGresultMap[“REPLY”].ToString（））；//响应消息
}
}
返回响应消息；
}
捕获（HttpRequestException e）
{
LogWriter.LogWrite（“错误：+e.Message”）；
System.Console.WriteLine（“错误：+e.Message”）；
返回null；
}
}
专用异步任务DoProcessInvocationStep（WaterCallStepContext stepContext，CancellationToken CancellationToken）
{
conversationData.index=0；//用于其他用途的变量
conversationData.result=等待
GetEPcallsDoneAsync（conversationData）；
Wait_conversationStateAccessor.SetAsync（stepContext.Context、conversationData、cancellationToken）；
返回Wait-stepContext.NextAsync（cancellationToken）；
}
公共类TopLevelDialog:ComponentDialog
{
专用只读IStatePropertyAccessor\u会话状态访问器；
会话数据会话数据；
公共顶级对话（会话状态会话状态）
：base（名称（TopLevelDialog））
{
_conversationStateAccessor=conversationState.CreateProperty（nameof（ConversationData））；
AddDialog（新建文本提示（名称）（文本提示））；
AddDialog（新建ChoicePrompt（名称）（ChoicePrompt））；
AddDialog（新建ReviewSelectionDialog（会话状态））；
AddDialog（新建EsSelectDialog（））；
AddDialog（新建WaterWallDialog）（名称（WaterWallDialog），新建WaterWallStep[]
{
StartSelectionStepAsync，
GetESResultStep，
doprocess步骤，
ResultStepAsync，
迭代步骤异步
}));
InitialDialogId=nameof（WaterWallDialog）；
}
专用异步任务StartSelectStepAsync（WaterWallStepContext stepContext，CancellationToken CancellationToken）
{
conversationData=Wait_conversationStateAccessor.GetAsync（stepContext.Context，（）=>new conversationData（））；
//功能代码
Wait_conversationStateAccessor.SetAsync（stepContext.Context、conversationData、cancellationToken）；
返回Wait-stepContext.NextAsync（null，cancellationToken）；
}
//其他对话框步骤
}
conversationDataconversationDataconversationDataUser AconversationDataUser BstartSelectsStepAsyncconversationData=wait\u conversationStateAccessor.GetAsync（stepContext.Context，（）=>new conversationData（）），他们只需为每个人重置conversationData
；conversationDataconversationData conversationData；用户A对话数据将为null/空
状态应该如何Sav
// Copyright (c) Microsoft Corporation. All rights reserved.
// Licensed under the MIT License.
const { Channels, MessageFactory } = require('botbuilder');
const {
AttachmentPrompt,
ChoiceFactory,
ChoicePrompt,
ComponentDialog,
ConfirmPrompt,
DialogSet,
DialogTurnStatus,
NumberPrompt,
TextPrompt,
WaterfallDialog
} = require('botbuilder-dialogs');
const { UserProfile } = require('../userProfile');
const ATTACHMENT_PROMPT = 'ATTACHMENT_PROMPT';
const CHOICE_PROMPT = 'CHOICE_PROMPT';
const CONFIRM_PROMPT = 'CONFIRM_PROMPT';
const NAME_PROMPT = 'NAME_PROMPT';
const NUMBER_PROMPT = 'NUMBER_PROMPT';
const USER_PROFILE = 'USER_PROFILE';
const WATERFALL_DIALOG = 'WATERFALL_DIALOG';
/**
* This is a "normal" dialog, where userState is stored properly using the accessor, this.userProfile.
* In this dialog example, we create the userProfile using the accessor in the first step, transportStep.
* We then pass prompt results through the remaining steps using step.values.
* In the final step, summaryStep, we save the userProfile using the accessor.
*/
class UserProfileDialogNormal extends ComponentDialog {
constructor(userState) {
super('userProfileDialogNormal');
this.userProfileAccessor = userState.createProperty(USER_PROFILE);
this.addDialog(new TextPrompt(NAME_PROMPT));
this.addDialog(new ChoicePrompt(CHOICE_PROMPT));
this.addDialog(new ConfirmPrompt(CONFIRM_PROMPT));
this.addDialog(new NumberPrompt(NUMBER_PROMPT, this.agePromptValidator));
this.addDialog(new AttachmentPrompt(ATTACHMENT_PROMPT, this.picturePromptValidator));
this.addDialog(new WaterfallDialog(WATERFALL_DIALOG, [
this.transportStep.bind(this),
this.nameStep.bind(this),
this.nameConfirmStep.bind(this),
this.ageStep.bind(this),
this.pictureStep.bind(this),
this.confirmStep.bind(this),
this.saveStep.bind(this)
]));
this.initialDialogId = WATERFALL_DIALOG;
}
/**
* The run method handles the incoming activity (in the form of a TurnContext) and passes it through the dialog system.
* If no dialog is active, it will start the default dialog.
* @param {*} turnContext
* @param {*} accessor
*/
async run(turnContext, accessor) {
const dialogSet = new DialogSet(accessor);
dialogSet.add(this);
const dialogContext = await dialogSet.createContext(turnContext);
const results = await dialogContext.continueDialog();
if (results.status === DialogTurnStatus.empty) {
await dialogContext.beginDialog(this.id);
}
}
async transportStep(step) {
// Get the userProfile if it exists, or create a new one if it doesn't.
const userProfile = await this.userProfileAccessor.get(step.context, new UserProfile());
// Pass the userProfile through step.values.
// This makes it so we don't have to call this.userProfileAccessor.get() in every step.
step.values.userProfile = userProfile;
// Skip this step if we already have the user's transport.
if (userProfile.transport) {
// ChoicePrompt results will show in the next step with step.result.value.
// Since we don't need to prompt, we can pass the ChoicePrompt result manually.
return await step.next({ value: userProfile.transport });
}
// WaterfallStep always finishes with the end of the Waterfall or with another dialog; here it is a Prompt Dialog.
// Running a prompt here means the next WaterfallStep will be run when the user's response is received.
return await step.prompt(CHOICE_PROMPT, {
prompt: 'Please enter your mode of transport.',
choices: ChoiceFactory.toChoices(['Car', 'Bus', 'Bicycle'])
});
}
async nameStep(step) {
// Retrieve the userProfile from step.values.
const userProfile = step.values.userProfile;
// Set the transport property of the userProfile.
userProfile.transport = step.result.value;
// Pass the userProfile through step.values.
// This makes it so we don't have to call this.userProfileAccessor.get() in every step.
step.values.userProfile = userProfile;
// Skip the prompt if we already have the user's name.
if (userProfile.name) {
// We pass in a skipped bool so we know whether or not to send messages in the next step.
return await step.next({ value: userProfile.name, skipped: true });
}
return await step.prompt(NAME_PROMPT, 'Please enter your name.');
}
async nameConfirmStep(step) {
// Retrieve the userProfile from step.values and set the name property
const userProfile = step.values.userProfile;
// If userState is working correctly, we'll have userProfile.transport from the previous step.
if (!userProfile || !userProfile.transport) {
throw new Error(`transport property does not exist in userProfile.\nuserProfile:\n ${ JSON.stringify(userProfile) }`);
}
// Text prompt results normally end up in step.result, but if we skipped the prompt, it will be in step.result.value.
userProfile.name = step.result.value || step.result;
// step.values.userProfile.name is already set by reference, so there's no need to set it again to pass it to the next step.
// We can send messages to the user at any point in the WaterfallStep. Only do this if we didn't skip the prompt.
if (!step.result.skipped) {
await step.context.sendActivity(`Thanks ${ step.result }.`);
}
// WaterfallStep always finishes with the end of the Waterfall or with another dialog; here it is a Prompt Dialog.
// Skip the prompt if we already have the user's age.
if (userProfile.age) {
return await step.next('yes');
}
return await step.prompt(CONFIRM_PROMPT, 'Do you want to give your age?', ['yes', 'no']);
}
async ageStep(step) {
// Retrieve the userProfile from step.values
const userProfile = step.values.userProfile;
// If userState is working correctly, we'll have userProfile.name from the previous step.
if (!userProfile || !userProfile.name) {
throw new Error(`name property does not exist in userProfile.\nuserProfile:\n ${ JSON.stringify(userProfile) }`);
}
// Skip the prompt if we already have the user's age.
if (userProfile.age) {
// We pass in a skipped bool so we know whether or not to send messages in the next step.
return await step.next({ value: userProfile.age, skipped: true });
}
if (step.result) {
// User said "yes" so we will be prompting for the age.
// WaterfallStep always finishes with the end of the Waterfall or with another dialog; here it is a Prompt Dialog.
const promptOptions = { prompt: 'Please enter your age.', retryPrompt: 'The value entered must be greater than 0 and less than 150.' };
return await step.prompt(NUMBER_PROMPT, promptOptions);
} else {
// User said "no" so we will skip the next step. Give -1 as the age.
return await step.next(-1);
}
}
async pictureStep(step) {
// Retrieve the userProfile from step.values and set the age property
const userProfile = step.values.userProfile;
// We didn't set any additional properties on userProfile in the previous step, so no need to check for them here.
// Confirm prompt results normally end up in step.result, but if we skipped the prompt, it will be in step.result.value.
userProfile.age = step.result.value || step.result;
// step.values.userProfile.age is already set by reference, so there's no need to set it again to pass it to the next step.
if (!step.result.skipped) {
const msg = userProfile.age === -1 ? 'No age given.' : `I have your age as ${ userProfile.age }.`;
// We can send messages to the user at any point in the WaterfallStep. Only send it if we didn't skip the prompt.
await step.context.sendActivity(msg);
}
// Skip the prompt if we already have the user's picture.
if (userProfile.picture) {
return await step.next(userProfile.picture);
}
if (step.context.activity.channelId === Channels.msteams) {
// This attachment prompt example is not designed to work for Teams attachments, so skip it in this case
await step.context.sendActivity('Skipping attachment prompt in Teams channel...');
return await step.next(undefined);
} else {
// WaterfallStep always finishes with the end of the Waterfall or with another dialog; here it is a Prompt Dialog.
var promptOptions = {
prompt: 'Please attach a profile picture (or type any message to skip).',
retryPrompt: 'The attachment must be a jpeg/png image file.'
};
return await step.prompt(ATTACHMENT_PROMPT, promptOptions);
}
}
async confirmStep(step) {
// Retrieve the userProfile from step.values and set the picture property
const userProfile = step.values.userProfile;
// If userState is working correctly, we'll have userProfile.age from the previous step.
if (!userProfile || !userProfile.age) {
throw new Error(`age property does not exist in userProfile.\nuserProfile:\n ${ JSON.stringify(userProfile) }`);
}
userProfile.picture = (step.result && typeof step.result === 'object' && step.result[0]) || 'no picture provided';
// step.values.userProfile.picture is already set by reference, so there's no need to set it again to pass it to the next step.
let msg = `I have your mode of transport as ${ userProfile.transport } and your name as ${ userProfile.name }`;
if (userProfile.age !== -1) {
msg += ` and your age as ${ userProfile.age }`;
}
msg += '.';
await step.context.sendActivity(msg);
if (userProfile.picture && userProfile.picture !== 'no picture provided') {
try {
await step.context.sendActivity(MessageFactory.attachment(userProfile.picture, 'This is your profile picture.'));
} catch (err) {
await step.context.sendActivity('A profile picture was saved but could not be displayed here.');
}
}
// WaterfallStep always finishes with the end of the Waterfall or with another dialog; here it is a Prompt Dialog.
return await step.prompt(CONFIRM_PROMPT, { prompt: 'Would you like me to save this information?' });
}
async saveStep(step) {
if (step.result) {
// Get the current profile object from user state.
const userProfile = step.values.userProfile;
// Save the userProfile to userState.
await this.userProfileAccessor.set(step.context, userProfile);
await step.context.sendActivity('User Profile Saved.');
} else {
// Ensure the userProfile is cleared
await this.userProfileAccessor.set(step.context, {});
await step.context.sendActivity('Thanks. Your profile will not be kept.');
}
// WaterfallStep always finishes with the end of the Waterfall or with another dialog; here it is the end.
return await step.endDialog();
}
async agePromptValidator(promptContext) {
// This condition is our validation rule. You can also change the value at this point.
return promptContext.recognized.succeeded && promptContext.recognized.value > 0 && promptContext.recognized.value < 150;
}
async picturePromptValidator(promptContext) {
if (promptContext.recognized.succeeded) {
var attachments = promptContext.recognized.value;
var validImages = [];
attachments.forEach(attachment => {
if (attachment.contentType === 'image/jpeg' || attachment.contentType === 'image/png') {
validImages.push(attachment);
}
});
promptContext.recognized.value = validImages;
// If none of the attachments are valid images, the retry prompt should be sent.
return !!validImages.length;
} else {
await promptContext.context.sendActivity('No attachments received. Proceeding without a profile picture...');
// We can return true from a validator function even if Recognized.Succeeded is false.
return true;
}
}
}
module.exports.UserProfileDialogNormal = UserProfileDialogNormal;