Archilogic has a solution that does everything you're describing above, and eliminates the cumbersome QGIS process. Here's an example:
https://archilogic-route-path.3d.io/

This is an app built as a prototype for benchmarking warehouse effectiveness that demonstrates:
- Using Mapbox to load two layers: (1) Raster SVG (converted to PNG in real time) of the base floor plan image, and (2) GeoJSON for the floor plan + asset polygons
- Determining an optimized route using an open source JavaScript route path library, and drawing it on the floor plan in Mapbox
Archilogic has a web-based editor that makes it easy to modify the underlying representation of space, as well as a conversion service offering that takes a PDF/image of a floor plan and converts it to a data model that is then accessible via SDK and API (in this example above, our REST API serves both the SVG and the GeoJSON used to load the floor plan data).
In our experience, onboarding the data, editing the data, and visualizing the data all have to be done holistically - point solutions tend to fail or end up requiring a ton of overhead to implement (like the OpenLayers + QGIS flow above). Another way to say that is - generating the GeoJSON you need to do any of these flows is usually a nightmare, then keeping those GeoJSONs up to date is an additional nightmare, and finally implementing a solution that can do anything with those GeoJSONs can be a substantial amount of work. Archilogic solves all of these problems in one platform.
Once you have the digital representation of your space, implementing wayfinding or path optimization can be:
- As simple as storing path coordinates
- As complex as implementing any one of the many great libraries that work out-of-the-box with mapping platforms like OSM, Google Maps, and Mapbox that can do these calculations for you (like this: https://github.com/royhobbstn/geojson-dijkstra)