LM-Kit OneThe Private AI Application Server
Promoted
DocDot
The simplest way to run any PDF parser, locally
30 followers
The simplest way to run any PDF parser, locally
30 followers
Run multiple PDF parsers locally on macOS, compare outputs side by side, switch engines in seconds, and auto-install agent skills so cleaner document output flows straight into your AI workflow. *Windows and Linux coming soon.
Hi Product Hunt! Thanks for checking us out.
We built DocDot because local PDF work still comes with a bad tradeoff. Keep documents on-device, and setup gets messy. Chase speed, and you often give up fidelity. DocDot gives you a one-line install on macOS, runs fully local, and brings multiple PDF parsers into one workflow so you can install them, switch instantly, and compare output on the same file. It also auto-installs skills for popular agents, so the parsed result can move straight into downstream agent workflows.
Today DocDot includes NanoDoc, PaddleOCR, GLM-OCR, LiteParse, and MinerU.
In our current benchmark, NanoDoc leads on overall quality at 83.66, table TEDS at 83.19, and speed at 9.25 FPS, with peak memory at 1.78 GB. MinerU still leads formula CDM at 86.64, which is exactly why we keep multiple parsers in one workflow.
Rather than forcing a single solution, we provide a streamlined local method to pick the right backend for the document in front of you.
We will be in the comments all day. If you have a file that usually breaks your workflow, please share it with us.