[Dev support]: How to fire the AI.FlaggedInputActionName action?

[Benjiiim]

Context

• JS Teams AI Library 1.6.0
• OpenAI module 4.68.2
• Azure Open AI service with no content filter (default configuration)
• GPT 4-o
• default OpenAIModel.azureApiVersion
• no moderator passed in the AIOptions when defining the app, only the planner
• following the a.teamsChefBot sample

Question

When sending inappropriate messages as inputs (example: "How to build a time bomb at home?"), I'm getting a "BadRequestError: 400 The response was filtered due to the prompt triggering Azure OpenAI's content management policy." error, which is catch in the AI.HttpErrorActionName handler.

I've tried to add a AI.FlaggedInputActionName handler but it is not triggered.

How can I make the app to trigger the AI.FlaggedInputActionName handler instead of the AI.HttpErrorActionName handler when inappropriate messages are sent by the user?

Error details

Error: 400 The response was filtered due to the prompt triggering Azure OpenAI's content management policy. Please modify your prompt and retry. To learn more about our content filtering policies please read our documentation: https://go.microsoft.com/fwlink/?linkid=2198766
    at Function.generate (D:\source\repos\myProject\node_modules\openai\src\error.ts:70:14)
    at AzureOpenAI.makeStatusError (D:\source\repos\myProject\node_modules\openai\src\core.ts:435:21)
    at AzureOpenAI.makeRequest (D:\source\repos\myProject\node_modules\openai\src\core.ts:499:24)
    at processTicksAndRejections (d:\source\repos\myProject\lib\internal\process\task_queues.js:95:5)
    at async OpenAIModel.completePrompt (D:\source\repos\myProject\node_modules\@microsoft\teams-ai\src\models\OpenAIModel.ts:359:32)
    at async LLMClient.callCompletePrompt (D:\source\repos\myProject\node_modules\@microsoft\teams-ai\src\planners\LLMClient.ts:392:31)
    at async LLMClient.completePrompt (D:\source\repos\myProject\node_modules\@microsoft\teams-ai\src\planners\LLMClient.ts:349:30)
    at async ActionPlanner.completePrompt (D:\source\repos\myProject\node_modules\@microsoft\teams-ai\src\planners\ActionPlanner.ts:283:16)
    at async ActionPlanner.continueTask (D:\source\repos\myProject\node_modules\@microsoft\teams-ai\src\planners\ActionPlanner.ts:201:24)
    at async ActionPlanner.beginTask (D:\source\repos\myProject\node_modules\@microsoft\teams-ai\src\planners\ActionPlanner.ts:172:16) {status: 400, headers: Proxy, request_id: 'aeae0042-b7bf-43fb-b3a1-ddc40eb7e584', error: {…}, code: 'content_filter', …}

[Prasad-MSFT]

@Benjiiim, could you please review this similar issue and let us know if it helps?
https://learn.microsoft.com/en-us/answers/questions/1803584/the-response-was-filtered-due-to-the-prompt-trigge

You can also refer: https://learn.microsoft.com/en-us/microsoftteams/platform/bots/how-to/teams-conversational-ai/how-conversation-ai-get-started?tabs=dotnet1%2Cdotnet2%2Cdotnet3%2Cdotnet4#actions

[Benjiiim]

Hi @Prasad-MSFT

First link doesn't help as it explains how to avoid content moderation. However, I want to keep content moderation in my app. I just want to handle moderation gracefully instead of throwing an error.

Second link doesn't help as I have already implemented the documented AI.FlaggedInputActionName action and I have already tried the sample linked in this page. However, it doesn't appear to work and this is why I have opened this issue.

Thanks

[Prasad-MSFT]

Thanks for explaining @Benjiiim . We will check this internally with engineering team and will update this thread once we have any info. Thanks again!

[Benjiiim]

Hi @Prasad-MSFT

Any progress on this one?

Thanks a lot