Quick Chat with Quick Dashboard with SDK

Hello,

I was under the impression with Quick Chat and Dashboards, we could create a chat agent that would know about the current dashboard we are showing. These are both being loaded by the SDK github.com/awslabs/amazon-quicksight-embedding-sdk.

However, it seems we can not pass in the a dashboard id to the Quick Chat.

I do understand that we can create an Quick agent and then attach a space with the dashboard, which does seem to work, however we have hundreds of dashboards, which seems silly to apply a Quick agent for each dashboard.

What is the correct solution for this?

Hi @jkubicki

Quick Chat agents (the QuickChat experience type in the SDK) are designed to work with Spaces and knowledge bases, they do not accept a dashboard ID or have direct dashboard awareness. As you noted, creating a separate agent per dashboard does not scale well for your use case.

The dashboard-aware conversational experience you’re looking for is the Generative Q&A embedding experience, a separate experience type available via the same GenerateEmbedUrlForRegisteredUser API.

Recommended approach: Generative Q&A with InitialTopicId

  1. Create QuickSight Topics that map to your underlying datasets. Since Topics are tied to datasets (not dashboards), multiple dashboards sharing the same dataset can share a single Topic. This means you likely need far fewer Topics than dashboards.

  2. Embed the Generative Q&A experience and dynamically pass the InitialTopicId based on which dashboard the user is currently viewing.

  3. Maintain a simple mapping in your application dashboardId -> topicId, and use it at URL generation time.

**Alternative: Console embedding with Q&A enabled: ** If you’d prefer the Q&A experience built directly into the dashboard view (rather than as a separate panel), you can use console embedding with the InitialPath set to your dashboard, and enable the DataQnA feature.

This approach embeds the full console scoped to a specific dashboard, with the AI-powered Q&A accessible within the same frame. Note that this requires Author Pro or Reader Pro roles for your users.

Please refer to the following official documentation, which might be helpful.

I see, to be honest Generative Q&A is also not what we are looking for.

While we can get information from the dataset with a Topic it is still disjointed from the dashboard we are displaying.

Ideally we want something that can describe in more detail what the set values or how to interpret information on the dashboard, not just the dataset.

For example if we change a value on the dashboard and the data looks like an anomaly, we would want to ask why that is happening. Maybe there is a better way of passing these changed parameters to the Topic so it understands what changed on the dashboard? Maybe we need to pass those event triggers to the Q&A?

It seems like the Quick Chat technically has that ability but it’s just not being exposed this way with the SDK. If there any future plans for this to be able to pass specific parameters like dashboard ids to the Quick Chat? Is there any way to see roadmap items for the Quick Chat SDK?

Hi @jkubicki

Apologies for the delayed response

Thank you for the clarification. I understand now that you’re looking for a conversational experience that is aware of the current dashboard state (active filters, parameters, and visuals) rather than just the underlying dataset. That’s an important distinction and a valid use case.

Within the native QuickSight UI, the default chat agent embedded in dashboards is scoped to the dashboard’s current filter state and visuals. However, this dashboard-aware context is not currently exposed through the embedding SDK’s QuickChat experience type.

I’ve logged this as a feature request to promote visibility with the product team.