Feature Request: More Traditional Installation Process about oobabot HOT 8 CLOSED

I would also prefer a more transparent installation process. It makes it easier to tinker with things when you know where everything is stored. I do recognize the value in a simple process, so perhaps having both options available would work.

from oobabot.

note for future folks: the person who filed this task has the username of TheMe[capitol i]onGod, not "TheMelonGod" (with a lower case l).

I think the person with the "TheMeLonGod" username would appreciate it if we didn't ping him by accident.

I've fixed the comments above, just for future reference.

from oobabot.

Thanks for the clarification!

I've just added clearer instructions on how to run from a clone. Check them out here, and let me know if they help:
https://github.com/chrisrude/oobabot/blob/main/docs/CONTRIBUTING.md#development-environment

My preference is to have most folks run via the pip install process, which is why I haven't documented the dev process in the main readme. The main reason for this is that oobabot doesn't have great tests, so I've been reliant on just running it (and the helpful reports of folks here!) to uncover bugs in new code.

By having both a git-repo method and a pip install method, the folks who are more excited to try the latest can do so (with the understanding that sometimes things will break), while folks that just want it to work can use pip install, and hopefully run into issues less frequently.

from oobabot.

Thanks for the feedback! I'll take a look.

from oobabot.

@Skrownerve what would a more transparent installation process look like?

@TheMeIonGod I've added the ability to pass a --discord-token via the command line, which should be included in the 0.1.6 release (probably next week).

As you may well know there are security concerns with passing a secret on a command line, mainly that anyone else who is using the machine and can run ps will be able to see it. But a few people have struggled with setting the env var, and I know that many of us are just doing this on single-user systems where the security concern is mitigated for other reasons. So I hope this might help you get started once it's out.

from oobabot.

@TheMeIonGod @Skrownerve i found installation like this much better , it just need more openess
Like we can set number of tokens , temperature , other parameters etc
I think it's perfect and easy

from oobabot.

My thoughts -- I like the current python install process. I know it's pretty rough on Windows, but on osx and linux it's pretty standard, and there's lots of tooling to support it.

I do think there are folks who would like to run oobabot who aren't super-comfortable with that, or just don't want to spend the energy to figure it out. To address those folks, I would like to do one of:

• create a plugin to oobabooga
• create a docker container where everything's bundled
  Both of these would build on the existing python egg, not replace it.

The plugin idea has been suggested by others, and would probably be the easiest, as you have to setup oobabooga anyway, then it would be one click plus some UI config.

This project is still pretty young, though, and I want to spend some time iterating before trying to integrate it. Development is more costly the bigger the project is, so I think it makes sense to try and keep things as simple as possible for now, and then as the project feature set starts to stabilize, the look to expand.

from oobabot.

@chrisrude By "transparent", I just mean more easily accessible files. I'm not too familiar with Python, and most of the projects in AI I've played with have traditional setups where I can dig through the files. Like, even on Linux via WSL, when I install Oobabooga's web UI, I have a file structure I can browse and see every file and make tweaks if I want to (that I have to inevitably stash so I can update to latest because I'm not confident trying to merge others' work with my tweaks).

To put it simply, maybe I'm just not aware of how/where I could inspect the files installed by pip. As of now, I don't know if it's possible, and therefore your install process, while simple, is opaque to me. That being said, it's not bad how it is now. And once you add all the configurations you've been planning I probably wouldn't even need to go digging around anyway.

For an example of a familiar process to me, the one-click installers for Oobabooga's and AUTOMATIC1111/Vladmandic's web UIs seem ideal. They create a venv and install all dependencies in them somewhere I can clearly see and meddle with, and give me clear instructions on how to run it. I have no idea if it's possible to run oobabot from the folder that results from a git clone. I tried and failed.

from oobabot.