Improving Overpass user experience thanks to AI

I’ve just come across this interesting thread explaining how to easily generate Overpass queries using GPT, which seems like a user experience booster for a lot of people and lowering the entry level to the minimum :slight_smile:

https://twitter.com/lemonodor/status/1636849040548675584

From the comments I’ve also learned about ChatGeoPT

overpass-ai is doing something similar:

Prompt example:

Write an overpass query to highlight all public transport lines in Madrid, Spain

Result:

[out:json];
area[name="Madrid"]->.a;
(
  relation["type"="route"]["route"="bus"](area.a);
  relation["type"="route"]["route"="tram"](area.a);
);
out body;
>;
out skel qt;

There are some open source models being released in the pasts weeks. Maybe we can dream with a future where you can interact, search and query osm.org using natural language? :star_struck:

7 Likes

Very exciting! It’s also a great example of entirely open data allowing modern and exiting uses.

Adding this site to the list:

2 Likes

Another tool using AI and overpass-turbo

1 Like

I tried making a “GPT” for this purpose, too. It is sometimes able to make a proper query and get the results, but frequently asks for too large of a result, or creates a malformed query.

Actually it does work OK today. Not sure why it was broken last time: https://chat.openai.com/g/g-qremyqcEQ-osm-query-interpreter

image