index processes (only last step)

[acidjunk]

MVP index for last step

[timfdev]

For surf data this is too big as it contains a very large mapping of uuids. This is not only too big but also not semantically relevant. Maybe we can store it in a single jsonb field and do similarity or exact matches on it only.

[acidjunk]

i think you can skip the mapping.

Do you have an example?
Probably also ok to remove some stack trace info during indexes etc. We might need cleanup before index.

Probably only tree is needed and maybe semantic on top level fields (without descendants)