Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
APPLIES TO: SDK v4
Handling interruptions is an important aspect of a robust bot. Users won't always follow your defined conversation flow, step by step. They may try to ask a question in the middle of the process, or simply want to cancel it instead of completing it. This article describes some common ways to handle user interruptions in your bot.
Note
To build agents with your choice of AI services, orchestration, and knowledge, consider using the Microsoft 365 Agents SDK. The Agents SDK has support for C#, JavaScript or Python. You can learn more about the Agents SDK at aka.ms/agents. If you're looking for a SaaS-based agent platform, consider Microsoft Copilot Studio. If you have an existing bot built with the Bot Framework SDK, you can update your bot to the Agents SDK. You can review the core changes and updates at Bot Framework SDK to Agents SDK migration guidance. Support tickets for the Bot Framework SDK will no longer be serviced as of December 31, 2025.
Prerequisites
- Knowledge of bot basics, managing state, the dialogs library, and how to reuse dialogs.
- A copy of the core bot sample in C#, JavaScript, Java or Python.
The core bot sample uses Language Understanding (LUIS) to identify user intents; however, identifying user intent isn't the focus of this article. For information about identifying user intents, see Natural language understanding and Add natural language understanding to your bot.
Note
Language Understanding (LUIS) will be retired on 1 October 2025. Beginning 1 April 2023, you won't be able to create new LUIS resources. A newer version of language understanding is now available as part of Azure AI Language.
Conversational language understanding (CLU), a feature of Azure AI Language, is the updated version of LUIS. For more information about language understanding support in the Bot Framework SDK, see Natural language understanding.
About this sample
The sample used in this article models a flight booking bot that uses dialogs to get flight information from the user. At any time during the conversation with the bot, the user can issue help or cancel commands to cause an interruption. There are two types of interruptions handled:
- Turn level: Bypass processing at the turn level but leave the dialog on the stack with the information that was provided. In the next turn, continue from where the conversation left off.
- Dialog level: Cancel the processing completely, so the bot can start all over again.
Define and implement the interruption logic
First, define and implement the help and cancel interruptions.
To use dialogs, install the Microsoft.Bot.Builder.Dialogs NuGet package.
Dialogs\CancelAndHelpDialog.cs
Implement the CancelAndHelpDialog
class to handle user interruptions. The cancelable dialogs, BookingDialog
and DateResolverDialog
derive from this class.
Warning
It looks like the sample you are looking for has moved! Rest assured we are working on resolving this.
In the CancelAndHelpDialog
class, the OnContinueDialogAsync
method calls the InterruptAsync
method to check if the user has interrupted the normal flow. If the flow is interrupted, base class methods are called; otherwise, the return value from the InterruptAsync
is returned.
Warning
It looks like the sample you are looking for has moved! Rest assured we are working on resolving this.
If the user types "help", the InterruptAsync
method sends a message and then calls DialogTurnResult (DialogTurnStatus.Waiting)
to indicate that the dialog on top is waiting for a response from the user. In this way, the conversation flow is interrupted for a turn only, and the next turn continues from where the conversation left off.
If the user types "cancel", it calls CancelAllDialogsAsync
on its inner dialog context, which clears its dialog stack and causes it to exit with a canceled status and no result value. To the MainDialog
(shown later on), it will appear that the booking dialog ended and returned null, similar to when the user chooses not to confirm their booking.
Warning
It looks like the sample you are looking for has moved! Rest assured we are working on resolving this.
Check for interruptions each turn
Once the interrupt handling class is implemented, review what happens when this bot receives a new message from the user.
Dialogs\MainDialog.cs
As the new message activity arrives, the bot runs the MainDialog
. The MainDialog
prompts the user for what it can help with. And then it starts the BookingDialog
in the MainDialog.ActStepAsync
method, with a call to BeginDialogAsync
as shown below.
Warning
It looks like the sample you are looking for has moved! Rest assured we are working on resolving this.
Next, in the FinalStepAsync
method of the MainDialog
class, the booking dialog ended and the booking is considered to be complete or canceled.
Warning
It looks like the sample you are looking for has moved! Rest assured we are working on resolving this.
The code in BookingDialog
isn't shown here as it's not directly related to interruption handling. It's used to prompt users for booking details. You can find that code in Dialogs\BookingDialogs.cs.
Handle unexpected errors
The adapter's error handler handles any exceptions that weren't caught in the bot.
AdapterWithErrorHandler.cs
In the sample, the adapter's OnTurnError
handler receives any exceptions thrown by your bot's turn logic. If there's an exception thrown, the handler deletes the conversation state for the current conversation to prevent the bot from getting stuck in an error loop caused by being in a bad state.
Warning
It looks like the sample you are looking for has moved! Rest assured we are working on resolving this.
Register services
Startup.cs
Finally, in Startup.cs
, the bot is created as a transient, and on every turn, a new instance of the bot is created.
Warning
It looks like the sample you are looking for has moved! Rest assured we are working on resolving this.
For reference, here are the class definitions that are used in the call to create the bot above.
Warning
It looks like the sample you are looking for has moved! Rest assured we are working on resolving this.
Warning
It looks like the sample you are looking for has moved! Rest assured we are working on resolving this.
Warning
It looks like the sample you are looking for has moved! Rest assured we are working on resolving this.
Test the bot
- If you haven't done so already, install the Bot Framework Emulator.
- Run the sample locally on your machine.
- Start the Emulator, connect to your bot, and send messages as shown below.
Additional information
The 24.bot-authentication-msgraph sample in C#, JavaScript, Python, or Java shows how to handle a logout request. It uses a pattern similar to the one shown here for handling interruptions.
You should send a default response instead of doing nothing and leaving the user wondering what is going on. The default response should tell the user what commands the bot understands so the user can get back on track.
At any point in the turn, the turn context's responded property indicates whether the bot has sent a message to the user this turn. Before the turn ends, your bot should send some message to the user, even if it's a simple acknowledgment of their input.