The standard library and CLI for Cicero tasks and actions.
We use Tullia as a handy tool for running code in isolation. It consumes task definitions written in Nix, and runs the task and its dependencies.
The goal for this project is to enforce identical environments during development, CI, and when deployed.
The task and its dependencies run in nsjail by default. There is also
podman support available, but it's not as mature yet.
This allows better isolation and control than a pure nix shell and behaves similiar to sandboxes of Nix derivations.
That means that the task will only be able to see files in the current working directory, and may only use explicitly declared dependencies. It's also possible to disable networking, restrict resources like CPU and RAM, amongst other things.
This can also be helpful when working on a derivation that takes a lot of time to build, by invoking a compiler but retaining caches between runs.
nix profile install github:input-output-hk/tullia❯ tullia list
┌ tullia run
├─┬ build
│ └── bump
├─┬ bump
│ └── lint
├─┬ lint
│ └── tidy
└── tidy
❯ tullia run build
[✔] done build 38.055659394s
[✔] done bump 36.479661466s
[✔] done lint 10.892518424s
[✔] done tidy 5.069157358sTullia can be invoked with the --mode flag to change its output and some
runtime behaviour.
With cli, the output is rendered in a pretty fashion, keeping track of
the time each task execution takes and showing logs only in case of errors.
Passing verbose shows the inner workings of task execution.
Setting the mode to json is similar to verbose, but instead of
human-readable logs it outputs JSONL which is better suited for further
digesting the logs.
The passthrough mode is mostly useful for recursive invocations of Tullia. In
this mode Tullia will not output anything itself, but instead pass on
std{in,out,err} to the tasks it invokes and give control over the tty to
them. This should only be required in rare cases.