https://github.com/benmills/openai-markdown

https://github.com/benmills/openai-markdown

gpt-3 gpt-4 markdown obsidian-md obsidian-plugin openai

Last synced: about 1 month ago
JSON representation

• Host: GitHub
• URL: https://github.com/benmills/openai-markdown
• Owner: benmills
• Created: 2023-05-07T16:19:52.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2023-05-08T00:07:06.000Z (over 1 year ago)
• Last Synced: 2024-04-17T02:56:28.551Z (7 months ago)
• Topics: gpt-3, gpt-4, markdown, obsidian-md, obsidian-plugin, openai
• Language: TypeScript
• Homepage:
• Size: 127 KB
• Stars: 12
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# OpenAI Markdown

Chat with OpenAI models as a markdown file inside Obsidian desktop or mobile. [Download the latest release here](https://github.com/benmills/openai-markdown/releases/latest) and add it to your `.obsidian/plugins` folder to get started.



* Edit the entire conversation, not just your next message, making few-shot prompting easy.

* Store and organize your chats in your Obsidian vault

* Configure your OpenAI API key and use either GPT-3.5-turbo or GPT-4.

**Example: chat.md**

What is 1 + 2?

✨📜: The answer is 3 ••◦

## How it works

**Conversation syntax**

OpenAI Markdown prioritizes minimal syntax to ensure your files are as close to pure markdown as possible. Model responses are wrapped with `Icons` that can be set to any text string in the plugin settings. By default, a sequence of emojis are used (✨📜, ••◦). This means that all text not wrapped in the icons is considered a user message.

**Send file to OpenAI**

Run the command `Send current file as conversation to OpenAI model` to extract the conversation from the current file and send it to the configured OpenAI model to get a response. Consider binding this to a hotkey like cmd+enter.

*Note: Model responses are streamed to the file which can cause issues with scrolling while the model is responding*

**Reruns**

By default, if you send a file to OpenAI without adding a new message it will delete the previous model response and regenerate it.

**Delete last model response**

Run the command `Delete last response from model` Consider binding this to a hotkey like cmd+backspace.

## Risks

This is an early release and likely has bugs. Please be careful and consider the risks before uses:

* **Your OpenAI API key will be stored in your Obsidian vault. This may expose your API key to other plugins. Consider generating a specific API key to ensure you can easily rotate.**

* Conversation syntax is very basic and if the `Icon` sequence appears in the model's output or in a user message it will break the parser.

* API requests have minimal error handling. This may result in hanging requests or UI lag.

* OpenAI models are normally very good about outputting valid markdown, but there is little to no output sanitation.

## Installation

[Download the latest release here](https://github.com/benmills/openai-markdown/releases/latest) and add it to your `.obsidian/plugins` folder.

**Build it yourself**

Clone this repo inside your `/.obsidian/plugins` folder. Run `npm i` or `yarn` and then `npm run build` or `yarn build`. You will see a file called `main.js` in the root directory if it worked.

As the plugin stabilizes it will be published as a community plugin.

## Reach out

Find me [@benemills](https://twitter.com/benemills) on twitter and [@benmills](https://warpcast.com/benmills) on Farcaster.