You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Like when you see this, simply draw a polygon around the cluster of interest.
Then, in the CHAT and OLLAMA integrations one could reference the selected chunks with a new placeholder like POLYGON_SELECTION (extending existing options SEARCH_RESULT, FULL_TEXT) for LLM context (e.g. "What do these text snippets have in common?"). This would enable faster insights into the data than manually hovering over the points and trying to understand why there is a cluster.
From a technical point of view, this requires the integration of https://visgl.github.io/deck.gl-community/docs/modules/editable-layers/ (former nebula.gl) to draw a polygon, get the GeoJSON and spatially filter the x/y coordinates (in fact, it's lat lon atm), either with a small JS script or directly with turf.js (might be overkill for this simple task but works like a charm).
Performance-wise it might make sense to cache the filtered points for subsequent calls.
Before taking on this feature, it might be best to embrace the state-based logic from deck.gl (I'm was totally avoiding for now) in order to alter point radius and color (without having the recreate the whole plot). I wrote some better code here: https://github.com/do-me/deckgl-label-collision.
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
The main idea here is to be able to select points via a polygon on the dim map for cluster exploration as described here: https://twitter.com/DomeGIS/status/1786524989602066795
Like when you see this, simply draw a polygon around the cluster of interest.
Then, in the CHAT and OLLAMA integrations one could reference the selected chunks with a new placeholder like POLYGON_SELECTION (extending existing options SEARCH_RESULT, FULL_TEXT) for LLM context (e.g. "What do these text snippets have in common?"). This would enable faster insights into the data than manually hovering over the points and trying to understand why there is a cluster.
From a technical point of view, this requires the integration of https://visgl.github.io/deck.gl-community/docs/modules/editable-layers/ (former nebula.gl) to draw a polygon, get the GeoJSON and spatially filter the x/y coordinates (in fact, it's lat lon atm), either with a small JS script or directly with turf.js (might be overkill for this simple task but works like a charm).
Performance-wise it might make sense to cache the filtered points for subsequent calls.
Before taking on this feature, it might be best to embrace the state-based logic from deck.gl (I'm was totally avoiding for now) in order to alter point radius and color (without having the recreate the whole plot). I wrote some better code here: https://github.com/do-me/deckgl-label-collision.
Beta Was this translation helpful? Give feedback.
All reactions