Show HN: Mark CLI – use Markdown, images and links to interact with GPT4o (github.com)

4 points by ryanelston 2y ago ↗ HN

Mark allows you to leverage Markdown file/syntax when interacting with GPT models, this provides a natural and efficient way to incorporate links and images as context for the LLM.

### Key Features: - *In-Document Thread Building*: Facilitate a threaded conversation directly in your Markdown document. - *GPT Vision with Image Tags*: Utilize GPT-4's vision capabilities. - *Effective RAG Retrieval Using Links*: Simple and intuitive method to include relevant documents as context. - *Custom System Prompts*: Personalize interactions with custom system prompts.

### Why Did I Create Mark? As someone who spends a lot of time writing and managing content, I needed a tool that was both flexible and efficient. Proprietary writing apps that incorporate AI are too restrictive. Other open-source tools require scripting to use the latest GPT features. Mark was developed to solve these pain points using the simplicity and versatility of Markdown, allowing for a natural workflow that integrates smoothly with my everyday tasks.

### Why Your Feedback Matters Since `Mark` is brand new, I'm eager to hear your thoughts and suggestions. Your feedback will be invaluable in refining the tool and making it even more useful for everyone.

Feel free to try it out and share your experience!

Check out [github.com/relston/mark](https://github.com/relston/mark) and leave your feedback here or post an issue. Thank you!

Warm regards, Ryan Elston

5 comments

[ 3.5 ms ] story [ 27.1 ms ] thread

withinboredom 2y ago ↗

Doesn't this involve extra costs from using the API?

ryanelston 2y ago ↗

This tool itself is FOSS, but it is designed to work using your own OpenAI api API key, so that is a cost that goes to OpenAI.

However I did add an OPENAI_API_BASE_URL env var override option, which allows you to point to any LLM endpoints that conform to the OpenAI spec. I use this at work because we have a proxy service called Credal which has a drop-in replacement API that the company pays for (they handle company IP and security stuff for us).

I am not an expert but I believe there is a tool our there that provides OpenAI compatible endpoints for open source LLMs, so in theory you can use this tool without any charges at all. I haven't played around with that myself though.

nickfixit 2y ago ↗

danielmiessler.com/p/fabric-origin-story. A tool I've been using as well.

Yeah I like fabric! I have played around with it. Eventually want to look for opportunities to make see if I can have `mark` and `fabric` work well together.

mud_dauber 2y ago ↗

I installed mark to investigate turning lots (a couple of hundred) of local PDFs into study guides and/or summaries. All of the PDF links bombed, mostly due to UTF-8 decode errors - which is fine. I pdftotext'd a few and ran into RateLimit Errors.

Otherwise - very promising!