Conversation
Add public AI predictions page with map view and publish/retract flow for offline predictions
|
Thanks for the review. All changes have been made @kshitijrajsharma. WRT to this:
We all agreed to the current design before implementation. Having the card clickable will require a design for the information panel and we can do that. Let's double check with @omranlm for the way forward. |
|
when I view MS project from public AI prediction and try to see MS results .. it is downloading the mbtiles .. |
|
From user perspective When we select specific model ID 150 for example and then we run prediction request using a selected model in the menu like YOLO base model, when we submit the prediction request we need to save the base model used in this request ... and it should show the used model in the AI predictions public page Documented in details publicly here https://drive.google.com/file/d/1c1oiZfKQsYm8PoTsdMicIwugFboZGfwS/view?usp=sharing |
sure, no worries for this public release we can keep it as cards only ... then when we support multiple features, we would need to design a page to show different type of features and more meta data coming with the used model. such as solap panels tags, other rooftops .. seagrass ... etc. |
Bug fix: Mapswipe result is downloading mbtiles instead of opening results modal
This has been resolved. |


What Does This PR Do ?
This PR introduces a public-facing Public AI Predictions page and completes the publish/retract workflow for offline predictions.
On the public side, it adds a new
/ai-predictionsroute, exposes it under the “Explore Datasets → AI Predictions” navbar menu, and wires it through NuqsAdapterso search, ordering, layout, pagination and map state are all synced via query params. The page supports grid, list and data-table layouts, plus a map view that renders clustered centroids with MapLibre; clicking a map point selects the corresponding prediction.Screenshots
Screen.Recording.2026-03-23.at.08.52.32.mov
How To Test
https://fair-dev-spatialnode.vercel.app/