Hi, I am trying to have the QS chat agent to analyze a QS dashboard to pull only the best results out of the 30+ Global Store lanes (to determine the best performing lane), for each of the business metrics (e.g. GMS, Glance Views, etc).
The problem is that I need to edit dashboard filters to display each lane separately (there is a filter for total, but that doesn’t give the individual per-lane results). I tried copying the “Share this view” URL from each of the filtered views of the dashboard, and feeding them to the AI agent. But it seems the agent can only access the base dashboard view (without filters), even when I feed each of the filtered URLs.
Please see the attached screenshot. The agent told me that the above restriction is due to the QS API.
The agent suggested me to either import each of the filtered views onto excels, or take screenshots and then paste them on PPT files (since Spaces doesn’t scan PNG files, and the chat only allows up to 20 files). That is a very time consuming process- I spent the whole day taking screenshots and pasting them onto PPT. The excel option would be similar considering the QS dashboard is separated into multiple widgets (meaning I need to import 5-6 files separately and combine them to create an excel with the whole lane view including all metrics).
Can the team please improve the API integration so that I can just use the “Share this view” URLs to feed into the agent? That would be very powerful.
Welcome to the Quick Community! I’d like to ask a few questions and learn more about your use case. Have you created a topic for the dashboard? If your topic has a column/attribute that differentiate each lane, then your chat agent should be able to only return the specific lane (lane 1, lane 2, etc.) based on natural language. So you might not need “Share this view” filtered URL.
I worked with our BI team (creator of dashboard) to try your Topics suggestion. I was excited that the agent was able to provide some data via natural language prompts, but only about 20-30% of dashboard data was pulled. For the rest, I needed to provide screenshots, which worked better. Please see attachment for reference.
I ultimately had to break down the per lane dashboard view into into 4-5 screenshots (with 1-2 screenshots per view, the numbers were probably too small for the agent to scan appropriately- the results were much less reliable). Even with this approach I noticed <5% data inaccuracies, where in some cases completely different numbers were being pulled. I ultimately had to manually check all the data, as my task was for management reporting purposes.
While the QS capabilities are promising, it leaves much room for improvement. (I might as well have manually copied/ pasted everything from dashboard onto an excel to begin with, rather than taking all those screenshots, starting a different conversation after 20 files, merging all results, and in the end validating those)
I’m not sure what the optimal solution would be (improve OCR capabilities for screenshots; improve dataset access capabilities for topic queries; etc), but it can be a very powerful tool once my above exercise can be done right & fast. I feel that many other Amazonians are spending time doing similar tasks.
Just checking back in since we haven’t heard from you in a bit. The ability to have the chat agent filter a dashboard has been something that other community members have asked for. I think I heard on a user series the feature may be on the roadmap, but I cannot say for certain since I am not an AWS employee. With that said, I will leave the ‘feature-request’ tag attached to this post to help increase visibility to the Quick team.
If you still have any additional questions related to your initial post, feel free to share them. Otherwise, any update you’re able to provide within the next 3 business days would be helpful for the community.
I was continuing my efforts in a separate exercise, where I needed to compile filtered numbers. I noticed the QS data pull capabilities were much better if I generate a PDF of the filtered view (attached) and fed that to the agent, instead of providing screenshots, although it was still placing numbers in incorrect cells (e.g. where a data row contained blank cells, the data in adjacent cells were sometimes placed incorrectly).
It’s good that the dashboard filtering feature is on the roadmap, as that would eliminate many hours of manual work. I’ll be on the lookout for developments on this- it’ll be a life changer.
