Feature Request: Provision Base Images

[SC-CTS]

It would be really nice if it was possible to provision base VMs locally by providing a custom provision.sh with the ability to point to the custom base when launching vibe in a project. e.g.

It would make project setup easier and faster, otherwise every project needs to run a setup script once to get the necessary tools and settings installed.

Currently this can be done with copy-pasting the images from certain projects around, I guess, but an automatic way would be nice.

In extension it might be nice to be able to point to an http URL to get a base image when initializing a project so they could be shared.

[lynaghk]

Yeah, I considered this as well.

Currently this can be done with copy-pasting the images from certain projects around, I guess, but an automatic way would be nice.

This is the crux --- is there a design of flags/commands for the CLI that can do this in a way that's obviously clearer and easier to learn than just a shell script on top of vibe?

Do you a specific workflow or design idea we can discuss?

[SC-CTS]

I do have an idea to suggest.
I could imagine adding a couple of commands:

• it (or run if something less fun is preferred): What the current invocation would do, launch the VM. As an optional argument pass the name of (or URL to) the base-image to use. I'm unsure if it should support the script, send, expect options; For neatness those might be moved to an init command. Then the base-image name would move to that as well.
• build: Would expect a name for the image it outputs. Options would be same as now except no-default-mounts as it would not have any default mounts. The output would be the provisioned base-image in ~/.cache/vibe/.

The workflow I would imagine:

Prepare a base-image for e.g. Java by running vibe build --script ./provision.sh jdk25 which would setup the JDK and in my case also include the aditional CA-Certs & maven mirror configs I need.

Then in one of my Java projects I would run vibe it jdk25 (and just vibe it in subsequent calls) to launch into the VM.

[lynaghk]

I'm probably going to stew on this for a while.

Two main requirements:

• Avoid adding bunch of new commands and CLI flags
• Make the behavior of any new commands and flags obvious in the three possible situations:
  • Running vibe for the first time on a computer (when the default.raw image hasn't been created and configured yet)
  • Running vibe for the first time in a directory (where .vibe/instance.raw doesn't exist)
  • Running vibe again in a directory (where .vibe/instance.raw does exist)

One potential idea is to add a --from flag which specifies the image to copy to the --to flag.
These would default to ~/.cache/vibe/default.raw and .vibe/instance.raw, respectively, giving the current behavior of vibe.

This would be equivalent to

cp [from] [to]
vibe [to]

so I'm not sure if it's really worth it.
It also raises the question of what should happen if to already exists --- do we error out (so you don't accidentally remove your old VM) or do we overwrite (so that the command is idempotent)?