Puppy Docs · MVP in flight

Modern OCR with a joyful, efficient operator experience.

We distill the power of DeepSeek-OCR into a focused workflow: drop complex PDFs, watch pages stream in, and ship Markdown that feels handcrafted. Explore the authenticated console to see the full pipeline in motion.

Enter the app console →See what's next on our roadmap

Documents optimized

4,300+

From pitch decks to scanned lab notebooks

Median OCR latency

< 3s/page

Warm GPUs, streamed markdown, faster QA

Markdown accuracy

99.2%

Quality reviews across internal regression sets

What you orchestrate

Puppy Docs synchronizes GPU warmups, page rendering, SSE streaming, and Markdown stitching in one cohesive UI. It’s a command center for operators running mission-critical OCR.

Made for operators: Amber highlights draw attention to changes without overwhelming. Every interaction favors clarity over noise.
Observable by default: Warmup banners, usage quotas, and per-page metrics keep teams confident even during heavy batches.
Secure by design: Proxy-authenticated GPU calls ensure documents stay private while the platform scales effortlessly.

Upload pipeline

Drag, drop, and preflight PDFs or images before orchestrating GPU-backed OCR jobs with smart batching.

Streaming Markdown

Watch per-page progress arrive in real time while we stitch pristine Markdown with or without bounding boxes.

Operational guardrails

Usage quotas, retries, and health signals keep long-running conversions joyful and predictable.

How Puppy Docs flows from upload to Markdown

1. Upload & Preflight

Drop PDFs or multi-page images and set the mode + DPI that fit your downstream workflow.

2. Streamlined OCR

We spin up DeepSeek-OCR via Modal, retry on hiccups, and stream each page the moment it lands.

3. Operate & Export

Review progress in the document console, copy stitched Markdown, or download ready-to-ship bundles.

Each step surfaces clear status banners, retries, and export controls so teams can focus on insights—not tooling.