PDF2MD – Rust+Redis+ClickHouse+VLLM conversion pipeline for PDFs
If you just want to use it, try here - https://pdf2md.trieve.ai . I think the LLM's are astoundingly good at converting complex powerpoint style infographics. I wouldn't normally think folks on HN would find this interesting as the general concept has been posted about already in the past few months. We were heavily inspired by Zerox[1]. However, the stack we went with was fun and over-engineered which is more likely to create interesting discussion. We use all the same tools at Trieve (our main product), but wanted to see if they would be a good fit for something that needed to…
In plain words
PDF2MD is a tool that converts PDF documents to markdown format using a pipeline built with Rust, Redis, ClickHouse, and VLLM. It uses large language models to process PDFs, particularly excelling at converting complex infographics and PowerPoint-style layouts into structured markdown. The tool is available as a web service and built entirely with open-source dependencies, completed in two weeks as a practical demonstration of the technology stack used by Trieve, its creators.
written from the facts on this page · September 2026
From the sources
In the maker’s words, at launch
If you just want to use it, try here - https://pdf2md.trieve.ai . I think the LLM's are astoundingly good at converting complex powerpoint style infographics. I wouldn't normally think folks on HN would find this interesting as the general concept has been posted about already in the past few months. We were heavily inspired by Zerox[1]. However, the stack we went with was fun and over-engineered which is more likely to create interesting discussion. We use all the same tools at Trieve (our main product), but wanted to see if they would be a good fit for something that needed to get built in a tighter timeline and we think they were! Took us 2 weeks to get this setup end-to-end and it's by no means complete (see roadmap in linked README). However, it's cool that a relatively cookie cutter web service like this can be created with pure open-source dependencies and non-standard Rust tooling so quickly. Rust won't kill your startup! - Minijinja templates for the UI[2] - PDFObject for doc display in-browser[3] - actix/actix-web HTTP server framework[4] - Redis queue macro for worker async processing[5] - Clickhouse for task storage[6] - chm CLI to handle Clickhouse migrations[7] - MinIO S3 for object storage[8] [1]: https://news.ycombinator.com/item?id=41048194 [2]: https://github.com/mitsuhiko/minijinja [3]: https://github.com/pipwerks/pdfobject [4]: https://github.com/actix/actix-web [5]: https://github.com/devflowinc/trieve/blob/main/pdf2md/server... [6]: https://github.com/ClickHouse/ClickHouse [7]: https://docs.rs/chm/latest/chm/index.html [8]: https://github.com/minio/minio
More ai this month
the category →
I trained a 125M-parameter transformer to autocomplete piano performances in real time (~108 notes/sec on an iPhone 15). The idea is basically GitHub Copilot or Tabnine, except instead of prompting it with code, you prompt it by playing a few notes on a MIDI piano. The model then continues what you played, entirely on-device. The app is free if anyone wants to try it. Happy to answer questions about the model, training, Core ML, or the many things that didn't work.
AI · 17d ago · simedw.com
Astute▲585Automate your B2B brand going viral, with new media creators
AI · 18d ago · company-app.joinastute.com
Hey HN, Henry from Cactus here! We previously released Cactus Needle, a 14MB agentic LLM for tool call, device use, and structured extraction for phones, wearables, smart homes, small robots and microcontrollers. We got really great feedback here, and have now incorporated the suggestions to release Needle 2. The whole model is a single 14MB binary that runs a full session in 28MB of RAM; 45m parameters at 2bit compression. Needle hits 500 tokens/sec decode speed on a Raspberry Pi 5, sits between 400-1,500 tokens/sec on VR devices like Meta Quest 3S and Apple Vision Pro, and ranges…
AI · 27d ago · cactuscompute.com
Launched alongside, November 2024
the whole month →
- IB
I wasn't quite sure if this qualified as "Show HN" given you can't really download it and try it out. However, dang said[0]: > If it's hardware or something that's not so easy to try out over the internet, find a different way to show how it actually works—a video, for example, or a detailed post with photos. Hopefully I did that? Additionally, I've put code and a detailed guide for the netboot computer management setup on GitHub: https://github.com/kentonv/lanparty Anyway, if this shouldn't have been Show HN, I apologize! [0]…
Dev tools · 2024 · lanparty.house
