Nikhil Singh [said]:
image.png
This is currently a dirty commit in my local repo. I just wanted to confirm that this is the expected output—
I mean,yada, yadawill be replaced (ignore that) XD.
https://github.com/juspay/omnix/pull/415
Regarding debug, one way to tackle is to make process-compose-flake to output a schemas output (see here) hardcoded for that specific process-compose configuration. And then om show can merge that flake's schemas output.
@Srid just tested on a new cloned juspay/services-flakes/llm, working fine.
can you test if allSystems is exposed in nix repl after you enable debug flag?
did a flake update in llm too and also enabled debug = true.
about juspay/vira
evalChecks.isValidProcess =
definition ? settings;
changing the evalChecks.isValidProcess to the above snippet in the omnix/flake-schemas will get it working.
below is a working screenshot
image.png
one way to tackle is to make
process-compose-flaketo output aschemasoutput.
Yes, this was my thought too. If it is a custom flake with a non standard output it should expose it's own schema.
Nikhil Singh said:
one way to tackle is to make
process-compose-flaketo output aschemasoutput.Yes, this was my thought too. If it is a custom flake with a non standard output it should expose it's own schema.
Go ahead with this, by making https://github.com/Platonic-Systems/process-compose-flake export the schema automatically which we can then use.
image.png
@Srid Screenshot is with newly spined up nix os in virtual.
please share nix --version as i am unable to reproduce this.
@Shivaraj B H can you please try nix --no-accept-flake-config --refresh run github:niksingh710/omnix/arbitrary-flake-schemas show . in juspay/services-flake/example/llm. make sure to enable debug flag for llm services flake.
You need to add aarch64-darwin to the list, as I'm on M1 mac.
That's another issue that needs to be resolved as well; systems should not be hardcoded like this.
@Srid
Go ahead with this, by making https://github.com/Platonic-Systems/process-compose-flake export the schema automatically which we can then use.
By this you meant me to integrate the schema with debug one to be added in process-compose?
We can make process-compose-flake output a schema but i don't feel like with debug it is the correct way to solve it.
as in https://determinate.systems/posts/flake-schemas/
it states that flake-schemas are there to enumerate and check the content of flake *outputs*.
but in case of process-compose-flake, it doesn't expose itself via flake output. it is exposed by perSystem Option of flake-parts.
Imo the right way to solve this will be making process-compose also expose a flake output named process-compose under which the perSystem definition can exist.
it should be something like process-compose.${system}.default = {} but right now it is like ${system}.process-compose.default in the repl.
You don't need debug, because when you generate the schemas it already has access to config anyway. If that's not viable for any reason, you can explore alternate ideas - for eg., perhaps something in meta attribute of the packages?
A big part of this task involves research and experimenting :-)
Last updated: Oct 20 2025 at 20:20 UTC