DEDUPED Hot AI repos, turned into use cases Get the Monday 5

How to host a private OXR parsing API on your own GPU with OXR

Run the OXR parsing service in Docker so your documents stay on your own hardware.

Short answer: git clone https://github.com/SoMarkAI/OXR.git; then hf download SoMarkAI/OXR-1.0 --local-dir ./models/OXR-1.0; docker build -f docker/Dockerfile.allinone -t oxr-1.0. All steps.

Who it is forAn engineering team with an NVIDIA GPU server that needs document parsing behind its own network.

The problemSending internal documents to a public demo API is not acceptable, and the free API is not for production.

The outcomeA local parsing API on port 8000 with a demo page and API docs, usable by your internal tools.

What you need first

  • A PDF or document image to parse
  • Python 3.12 or Node.js for the SoMark CLI, or Docker for a local service
  • For local hosting: Linux, an NVIDIA GPU with BF16 support, and 24GB or more of GPU memory

Steps

  1. git clone https://github.com/SoMarkAI/OXR.git
  2. hf download SoMarkAI/OXR-1.0 --local-dir ./models/OXR-1.0
  3. docker build -f docker/Dockerfile.allinone -t oxr-1.0 .
  4. docker run --rm --gpus all --ipc=host -p 127.0.0.1:8000:8000 -v "$PWD/models/OXR-1.0:/models/oxr-model:ro" oxr-1.0 (inferred, check the README)
  5. Open http://localhost:8000/docs once the startup logs show the services are ready, then send a file to /v1/parse/sync.

Prompt for your agent

Using the OXR repo at https://github.com/SoMarkAI/OXR, download the OXR-1.0 weights, build the allinone Docker image, and start it on port 8000. Verify that http://localhost:8000/docs loads, then parse <path-to-document.pdf> via POST to /v1/parse/sync and save the returned Markdown.

Caveat. The README lists 24GB of GPU memory as the minimum for low-concurrency testing and recommends 48GB or more for long documents, and the first startup may take longer than usual.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for SoMarkAI/OXR and "Host a private OXR parsing API on your own GPU"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer