Home

Awesome

Star on GitHub Package Control

OpenAI Sublime Text Plugin

tldr;

Cursor level of AI assistance for Sublime Text. I mean it.

Works with all OpenAI'ish API: llama.cpp server, ollama or whatever third party LLM hosting.

Features

ChatGPT completion demo

https://github.com/yaroslavyaroslav/OpenAI-sublime-text/assets/16612247/37b98cc2-e9cd-46a6-ac5d-03845313096b

video sped up to 1.7x

https://github.com/yaroslavyaroslav/OpenAI-sublime-text/assets/16612247/69f609f3-336d-48e8-a574-3cb7fda5822c

video sped up to 1.7x

Requirements

Installation

  1. Install the Sublime Text Package Control plugin if you haven't done this before.
  2. Open the command palette and type Package Control: Install Package.
  3. Type OpenAI and press Enter.

Usage

AI Assistance use case

ChatGPT mode works the following way:

  1. Select some text or even the whole tabs to include them in request
  2. Run either OpenAI: Chat Model Select or OpenAI: Chat Model Select With Tabs commands.
  3. Input a request in input window if any.
  4. The model will print a response in output panel by default, but you can switch that to a separate tab with OpenAI: Open in Tab.
  5. To get an existing chat in a new window run OpenAI: Refresh Chat.
  6. To reset history OpenAI: Reset Chat History command to rescue.

[!NOTE] You suggested to bind at least OpenAI: New Message, OpenAI: Chat Model Select and OpenAI: Show output panel in sake for convenience, you can do that in plugin settings.

Chat history management

You can separate a chat history and assistant settings for a given project by appending the following snippet to its settings:

{
    "settings": {
        "ai_assistant": {
            "cache_prefix": "your_project_name"
        }
    }
}

Additional request context management

You can add a few things to your request:

To perform the former just select something within an active view and initiate the request this way without switching to another tab, selection would be added to a request as a preceding message (each selection chunk would be split by a new line).

To send the whole file(s) in advance to request you should super+button1 on them to make all tabs of them to become visible in a single view group and then run [New Message|Chat Model] with Sheets command as shown on the screen below. Pay attention, that in given example only README.md and 4.0.0.md will be sent to a server, but not a content of the AI chat.

[!NOTE] It's also doesn't matter whether the file persists on a disc or it's just a virtual buffer with a text in it, if they're selected, their content will be send either way.

Image handling

Image handle can be called by OpenAI: Handle Image command.

It expects an absolute path to image to be selected in a buffer or stored in clipboard on the command call (smth like /Users/username/Documents/Project/image.png). In addition command can be passed by input panel to proceed the image with special treatment. png and jpg images are only supported.

[!NOTE] Currently plugin expects the link or the list of links separated by a new line to be selected in buffer or stored in clipboard only.

In-buffer llm use case

Phantom use case

Phantom is the overlay UI placed inline in the editor view (see the picture below). It doesn't affects content of the view.

  1. You can set "prompt_mode": "phantom" for AI assistant in its settings.
  2. [optional] Select some text to pass in context in to manipulate with.
  3. Hit OpenAI: New Message or OpenAI: Chat Model Select and ask whatever you'd like in popup input pane.
  4. Phantom will appear below the cursor position or the beginning of the selection while the streaming LLM answer occurs.
  5. You can apply actions to the llm prompt, they're quite self descriptive and follows behavior deprecated in buffer commands.
  6. You can hit ctrl+c to stop prompting same as with in panel mode.

[!IMPORTANT] Yet this is a standalone mode, i.e. an existing chat history won't be sent to a server on a run.

[!NOTE] A more detailed manual, including various assistant configuration examples, can be found within the plugin settings.

[!WARNING] The following in buffer commands are deprecated and will be removed in 5.0 release.

  1. [DEPRECATED] You can pick one of the following modes: append, replace, insert. They're quite self-descriptive. They should be set up in assistant settings to take effect.
  2. [DEPRECATED] Select some text (they're useless otherwise) to manipulate with and hit OpenAI: New Message.
  3. [DEPRECATED] The plugin will response accordingly with appending, replacing or inserting some text.

Other features

Open Source models support (llama.cpp, ollama)

  1. Replace "url" setting of a given model to point to whatever host you're server running on (e.g."http://localhost:8080").
  2. [Optional] Provide a "token" if your provider required one. Temporarily mandatory, see warning below.
  3. Tweak "chat_model" to a model of your choice and you're set.

[!WARNING] Due to a known issue, a token value of 10 or more characters is currently required even for unsecured servers. More details here.

[!NOTE] You can set both url and token either global or on per assistant instance basis, thus being capable to freely switching between closed source and open sourced models within a single session.

Settings

The OpenAI Completion plugin has a settings file where you can set your OpenAI API key. This is required for the most of providers to work. To set your API key, open the settings within Preferences -> Package Settings -> OpenAI -> Settings and paste your API key in the token property, as follows:

{
    "token": "sk-your-token",
}

[!WARNING] Due to a known issue, a token value of 10 or more characters is currently required even for unsecured servers. More details here.

Advertisement disabling

To disable advertisement you have to add "advertisement": false line into an assistant setting where you wish it to be disabled.

Key bindings

You can bind keys for a given plugin command in Preferences -> Package Settings -> OpenAI -> Key Bindings. For example you can bind "New Message" including active tabs as context command like this:

{
    "keys": [ "super+k", "super+'" ],
    "command": "openai", // or "openai_panel"
    "args": { "files_included": true }
},

[Multi]Markdown syntax with syntax highlight support

It just works.

[!IMPORTANT] It's highly recommended to install the MultimarkdownEditing package to apply broader set of languages with syntax highlighting.

Proxy support

You can setup it up by overriding the proxy property in the OpenAI completion settings like follow:

"proxy": {
    "address": "127.0.0.1", // required
    "port": 9898, // required
    "username": "account",
    "password": "sOmEpAsSwOrD"
}

Known bugs

Workaround for #64

[!IMPORTANT] If any of your assistants are configured with a "token" of less than 10 characters, they will always fail to establish a connection, even if the server does not require authentication at all. Instead, the plugin will produce the following error message:

"No API token provided, you have to set the OpenAI token into the settings to make things work."

Luckily you'll find a quick and easy workaround here until a fix can be released!

Disclaimers

[!WARNING] All selected code will be sent to the OpenAI servers (if not using custom API provider) for processing, so make sure you have all necessary permissions to do so.

[!NOTE] This one was initially written at 80% by a GPT3.5 back then. I was there mostly for debugging purposes, rather than digging in into ST API. This is a pure magic, I swear!