add Agentic

[alecGraves]

• I'm the package's author and/or maintainer.
• I have have read the docs.
• I have tagged a release with a semver version number. (1.0.0)
• My package repo has a description and a README describing what it's for and how to use it.
• My package doesn't add context menu entries. *
• My package doesn't add key bindings. **
  • This package uses keybindings that only trigger on files created by the package for a minimal chat interface.
  • Scope of keybindings are limited using a "context" key that looks for file settings set by the plugin.
• Any commands are available via the command palette.
• Preferences and keybindings (if any) are listed in the menu and the command palette, and open in split view.
• If my package is a syntax it doesn't also add a color scheme. ***
• I use .gitattributes to exclude files from the package: images, test files, sublime-project/workspace.

My package is Agentic, a minimal OpenAI API v1/chat/completions interface for launching multiple custom LLM 'actions' on a file or code snippet in a project and streaming results.

My package is similar to others such as OpenAI completion however it should still be added because it focuses on achieving a minimal interface for rapidly launching repetitive user-defined custom agentic actions like 'simplify this code' or 'comment this code'. My package further focuses on streaming from multiple LLMs concurrently to enable multi-LLM workflows, and it supports configuration of pools of models for specific actions through package settings.

[alecGraves]

I fixed messages.json and renamed commands and custom document settings to all be prefixed with "agentic_..."
Thanks.

[alecGraves]

I also tested the plugin with sublime 3211 and it runs fine. I think Agentic should mostly work for any >= 3000. I am only using python 3.3 features and Python standard library, with basic sublime features for file manipulation.

[braver]

I think Agentic should mostly work for any >= 3000

That's fine. I would actually recommend opting in to the python 3.8 runtime if you can, and testing on python 3.13+ as that runtime version is just around the corner for ST. Future-proofing is more important that worrying about versions from more than 6 years ago 😄 People who don't upgrade their ST installation for that long probably aren't looking for the latest LLM implementation 😉

[braver]

Please check for prints to the console (e.g. https://github.com/alecGraves/Agentic/blob/master/chat_stream.py#L228) that can be removed (or put behind a debug mode).

It looks like your package would need quite a bit of setup to work, but with some reading through the readme and settings file I think I've got most of it figured out... It could perhaps use a getting started guide, since unless you're lucky and have models running at the exact URLs in your examples, not much will work out of the box.

Really interesting package though, nice 👍🏻

[alecGraves]

Thanks for the review. I appreciate the suggestions.
I made logging to console an optional setting, and I added a much-needed 'getting started' section to the readme.