chunkr

vision based document parsing

[chunkr @chunkrai]

PDFs suck.

If you want to build standout RAG apps, you need tooling that is tailored to your needs.

You should own your doc parsing infra.

Today we’re launching https://t.co/rm96AFicMJ on @ycombinator

https://t.co/MrEuBWBBvo

[chunkr @chunkrai]

chunkr-layout-1 and our other models are live now on our API!
🌐 Website: https://t.co/rm96AFicMJ
📷 Hugging Face Dataset: https://t.co/k1RQmt4yZf
📷Blog: https://t.co/V2BosKgY4n

[chunkr @chunkrai]

Today we're thrilled to announce the release of our newest layout-analysis model: chunkr-layout-1

Trained on millions of the hardest documents, across all verticals, chunkr-layout-1 is built to understand real documents, not just the clean ones.

[chunkr @chunkrai]

We can even parse floor plans/technical diagrams.

[Ishaan Kapoor @Ishaank1999]

Introducing chunkr-layout-1. Layout is where document intelligence starts, we just leveled it up.

- Benched on 1,013 hand-tagged samples.
80.9 mAP@50 | 88.4 P | 85.6 R | 86.9 F1

- Beats AWS, Azure, Docling, Gemini 2.5 Pro, OSS models

- Identifies 16 class labels

Live now!

[Ishaan Kapoor @Ishaank1999]

I’m thrilled to announce our new VLMs, chunkr-parse-1 and chunkr-parse-1-thinking.

- Parses complex forms, tables & more
- Inline OCR (redlining and formulas)
- Multilingual (>100 languages)
- Beats AWS Textract, Gemini 2.5 Pro, Mistral at OCR

Live now on the Chunkr API

[Philip Kung @philipkung]

Sweetspot does that: it can auto-detect fields (including from scanned PDFs) and fill out forms of 30+ pages using context from your organization

[chunkr @chunkrai]

webhooks, live now!

https://t.co/vM3xFy3zKe

[Ishaan Kapoor @Ishaank1999]

web hooks - live on @chunkrai now

chad @notakhilesh99 shipped it in like a day

docs below

[Ishaan Kapoor @Ishaank1999]

hello!

we are hiring @chunkrai

we are building best-in-class tools to help make the most out of documents. we're growing fast and are looking for a founding engineer to join!

if you

- like rust
- can handle features e2e
- are funny

lets talk!

https://t.co/kgVfcm6u00

[chunkr @chunkrai]

@skeptrune

[chunkr @chunkrai]

@m2oba resolved over email. small bug in reading order alg - should be fixed

[chunkr @chunkrai]

on weekends we bug bash.

the reception of our new excel parser has been amazing. after getting some feedback we're releasing a new, faster, more stable version.

enjoy!

[chunkr @chunkrai]

link to the cookbook here: https://t.co/1hJtFBSfuH

[chunkr @chunkrai]

Recently, we launched a parser that breaks up spreadsheets into clean python objects.

We used it to analyze the DoE’s 2020 Congressional Action Report. Here’s the data loaded via Chunkr vs openpyxl, huge difference.

[chunkr @chunkrai]

If you’ve ever tried to read an excel file into python, you’d know how cancerous it is.

It becomes really easy to analyze data when you're able to read spreadsheets cleanly.

[Nick Khami @skeptrune]

use chunkr

by far the most accurate and developer friendly product in the OCR document processing space

[Ishaan Kapoor @Ishaank1999]

Excels are notoriously hard to parse.

Even though they are machine readable, they end up as a flat string that loses semantic meaning of tables, styling and formulas. Agents can’t use it.

We’ve solved that.

Raw spreadsheets to perfect HTML, live now on the @chunkrai API.