Feature Suggestion: Markdown Resume Import

[korgano]

Currently, there is only one mode for importing resume data without generating it in the app itself.

PDFs are a common format that is generated by other resume tools, but it's understandable that the app can only import its own format.

Local AI users are likely to either have PDF to Markdown conversion capabilities for RAG, or can easily stand up such capabilities. This might be a good alternative to DOCX import with minimal requirements for post-process editing, especially for long resumes.

[gibbenergy]

Hi @korgano!

Great thought on PDF import! I really appreciate the suggestion.

Current Priority: UI/UX Experience

Right now, my main focus is on improving the UI/UX with Tailwind / React to make the app more intuitive and polished. I want to ensure users have a great experience with the core features before adding new capabilities.

Once the UX improvements are complete, I'll definitely revisit PDF import. It's on my radar and I'm keeping your suggestion open!

PDF Understanding: DeepSeek-OCR

For PDF extraction, I am thinking about the state-of-the-art DeepSeek-OCR (https://github.com/deepseek-ai/DeepSeek-OCR) model (released late 2025). It's specifically designed for OCR and document understanding with some impressive capabilities:

• 97% accuracy for structured documents
• 7-20x more efficient than traditional OCR (vision-text compression)
• 200K+ pages/day on a single GPU
• Purpose-built for complex layouts (multi-column, tables, etc.)

It seems like the perfect fit for resume parsing, especially with the complex layouts you mentioned.

What do you have in mind?

[korgano]

Hi @korgano!

Great thought on PDF import! I really appreciate the suggestion.

Current Priority: UI/UX Experience

Right now, my main focus is on improving the UI/UX with Tailwind / React to make the app more intuitive and polished. I want to ensure users have a great experience with the core features before adding new capabilities.

I have one major suggestion in that regard. Have more resume update buttons throughout the UI. It's not very user friendly for it to be only on the AI Resume Helper page, especially if you're doing a quick update of information on one of the pages.

PDF Understanding: DeepSeek-OCR

For PDF extraction, I am thinking about the state-of-the-art DeepSeek-OCR (https://github.com/deepseek-ai/DeepSeek-OCR) model (released late 2025). It's specifically designed for OCR and document understanding with some impressive capabilities:

* 97% accuracy for structured documents

* 7-20x more efficient than traditional OCR (vision-text compression)

* 200K+ pages/day  on a single GPU

* Purpose-built for complex layouts (multi-column, tables, etc.)

It seems like the perfect fit for resume parsing, especially with the complex layouts you mentioned.

Interesting. Hopefully it won't be too hard on the hard drive space.

[gibbenergy]

I just released v3, which focuses on modernizing the UI. The core logic remains the same, but the app now uses a more polished Tailwind based interface, and I implemented the UI suggestion around clearer update actions and buttons.

Regarding the second suggestion about PDF to Markdown, I agree this would be very useful, but it is currently non trivial. Resumes come in many formats, including two column layouts, varying text density, tables, and custom spacing. In practice, PDF to Markdown conversion is often inaccurate, and vision based models can hallucinate or misinterpret content when text is dense or poorly structured. That risk is especially high for resumes, where precision matters.

I think this is something worth revisiting later when the underlying tooling improves or when we can guarantee acceptable accuracy. For now, I prefer to avoid adding a feature that could silently degrade results.

[gibbenergy]

Currently, there is only one mode for importing resume data without generating it in the app itself.

PDFs are a common format that is generated by other resume tools, but it's understandable that the app can only import its own format.

Local AI users are likely to either have PDF to Markdown conversion capabilities for RAG, or can easily stand up such capabilities. This might be a good alternative to DOCX import with minimal requirements for post-process editing, especially for long resumes.

@korgano , I actually created a downstream Vision Ingest repo specifically to address this use case, handling resumes in different formats and converting them into Markdown and structured JSON.

Link: https://github.com/gibbenergy/Vision-Ingest

If you have time, I would really appreciate it if you could take a look and see whether it fits your workflow, especially for PDF based resumes. For most machine readable resumes the success rate has been good, while low quality scans or photos still require caution due to potential hallucination.

I will leave this issue open for a week or two to gather feedback, and then plan to close it once we decide whether this approach covers the import needs adequately.

Thanks again for the thoughtful input.

[korgano]

I'll give it a shot and provide feedback.

[korgano]

Ran into some problems with VisionParse. Do you want me to report them here, or in Vision Parse's repo?

[gibbenergy]

Please report them directly in that repo. I just cleaned it up and renamed the project to vision ingest, because there is already an existing repo named vision parse. If you open issues there, I can see them easily and try to fix them. Thanks a lot for testing this.