We don't really have much to support this. We have though about using
persistant_cache
plus wasm to make pre-computed applications.
basically for things that don't work in wasm (because of package limitations, certain network requests, size limits) you could pre-compute all variartions of how the use the app and then when users do interact in wasm, we would get all cache hits. and avoid re-running cells that may not work.
this approach could be really fragile and complex, so unsure if we will support it.
You could get pretty close to it yourself though if you'd like with wasm:
if is_wasm():
df = pl.read_csv("s3://my_cache//..)
else:
df = make_expensive_computation()
not sure if its only dataframes, but that could get you pretty far.
def is_wasm():
return 'pyodide' in sys.modules