Q: Are we able to embed a chat agent in the sidebar of the dashboard? I’ve only been able to successfully embed the conversational chat experience, but what I’d really like to be able to do is display my user in their dashboard with an agent co-pilot alongside to analyze the dashboard — like you can do in the console.
A: Right now the experience is what you see — you can embed it within the application and it sits alongside your dashboard. You will have to build an application like that to make it appear seamless, so it looks like it’s within your portal itself. The whole user experience would have the embedded dashboard and then the Quick Chat agent alongside it, as shown in the demo.
Q: Is it possible to make a template of the data sources of the chat? For example, if a client has different data sources, to roll it out they need to change the source of each client’s environment.
A: Currently, the chat points to a knowledge source or space. The way it works right now is that if you have a data source, you create a topic or a dashboard and enable it. To switch between data sources, you’d have to modify the dataset. If you need multiple data source datasets, you have to switch the chat agent on the fly — meaning you’d need multiple chat agents currently, because right now there are no APIs to flip the resource within the space. You can flip between chat agents using the SDK, but currently there is no way to flip the data source dynamically.
Q: What about adding some security guardrails? We have certain datasets that only a specific set of users are allowed to see. How do we pass that security context to the chat agent?
A: That is part of the configuration of the chat itself. When you’re creating the space or the knowledge base, if you have a dataset, you specify security there. For unstructured data, there are ACLs you can use. This has to be specified when you’re creating the space — giving access to the correct space for the correct users, and then passing it down to the dataset level or configuring it within the dataset itself.
Q: Are we able to organize chat agents and other objects (for example, spaces) in shared folders like you can with dashboards, datasets, and analyses?
A: Shared folders in Quick are currently for QuickSight assets — that is, all structured data. That is why Spaces were created. Spaces are for both structured and unstructured data. You can add topics, dashboards, and knowledge bases to a Space, but folders are only for structured data.
Q: If the chat is embedded in an app, does that user need to be a Quick user? Will it not load if they are not?
A: Yes. Embedded chat has launched only for registered users, so the user must be a registered user within Quick itself. Anonymous embedding is on the roadmap, but there are no timelines for it currently. Only registered users are supported at this time.
Q: Are we able to toggle on and off the chat history? There are specific use cases where we’d like to hide that option in the embedded container for specific personas.
A: The memory and conversation history is part of the UI and can be hidden. However, the updated SDK doesn’t currently have a hook to turn off history. This is good feedback — it will be taken back to the team to see if a hook can be added to turn off the history.
Q: Are there any upcoming API updates to support spaces, knowledge bases, or agent management?
A: Yes, the team is actively working on it. Stay tuned to the community — when those updates are available, they will be posted there.
Q: Can we add integrations to a space to share them?
A: Yes, absolutely.
Q: Is there a way to embed a similar or same agent across Quick Suite logins? For example, if there is relevant data across multiple accounts, can we apply a consistent agent for aggregation and analysis? Can we point the chat or a space to access knowledge and data from multiple accounts?
A: Not directly, because everything sits within a single account. If you need to access multiple accounts, you’d have to do something outside of the account. That said, if you have structured data sources and a data lake in a different account — for example, connecting via Athena — as long as you bring that dataset into the same QuickSight account, there are ways to connect to different accounts to access that source. However, the knowledge base will have to reside in the Quick account that the chat is accessing.