okraPDF field notes

Engineering documents for agents

Reproducible benchmarks, working code, and the failure modes behind PDF extraction. Written for people shipping document agents—not a content farm.

Browse the publication

Find the useful part

Showing all 21 articles

Benchmarks Aug 8, 2026

Local PDF parsing on a MacBook: where ParseBench says we actually are

ParseBench scores for every PDF parser that runs on a 16GB MacBook — Docling, Dots OCR, Qwen 3 VL, MinerU, Chandra — vs hosted APIs, plus a live tracker.

Agent engineering Aug 31, 2026

Claude vs Codex: How Two Agents Read PDFs by Default

Anthropic's and OpenAI's official PDF skills read side by side — text-first vs render-and-look, and what each one knows about forms.

Benchmarks Aug 18, 2026

Bill of lading OCR benchmark: 24 parsers, 25 BoLs, one leaderboard

24 PDF parsers scored on 25 bills of lading — digital, scanned, and fax-quality. Field-level accuracy and where each parser class breaks.

Extraction Sep 7, 2026

How to parse PDFs offline with Docling, Chandra OCR 2, and Ollama

Offline PDF parsing that works: Docling on CPU, granite-docling on Apple Silicon, Chandra OCR 2 on a GPU, Ollama-served models, real timings and failures.

Workflows Aug 28, 2026

How to track views on a PDF link

Share a PDF as a link and get honest view counts. What counts as a view, why bots and Range requests inflate numbers, and how to read the daily chart.

Extraction Aug 31, 2026

Extract Tables from PDF in Python: pdfplumber, Camelot & Tabula

A code-first guide to extracting tables from PDFs in Python — the open-source libraries that work, exactly where they break, and when to use an API.

Benchmarks Aug 31, 2026

Handwriting OCR benchmark: okraPDF reads IAM at 1.29% CER

okraPDF, Gemini 3, TrOCR, and Tesseract on the IAM handwriting benchmark, same lines and same scoring. Character error rates, methodology, and what they mean.

Agent engineering Aug 31, 2026

Generate PDFs in Code with okra exec (Codemode)

okra exec runs sandboxed JavaScript that composes okra's render engines — designed templates, ReportLab, HTML to PDF, and merge — and returns one PDF.

Workflows Aug 31, 2026

Extract invoice data from Gmail PDFs to Google Sheets with n8n

Build an n8n workflow that turns Gmail invoice PDFs into structured Google Sheets rows with okraPDF: upload, parse with a JSON schema, poll, append.

Benchmarks Jun 10, 2026

GDP.pdf: why the hardest PDF benchmark isn't really about PDFs

Surge AI's GDP.pdf holds every frontier model under 30%. What makes it hard: all-or-nothing rubrics, pre-encoded traps, and pixel-bound facts.

Extraction Aug 23, 2026

OCR for Scanned Documents: A Production Pipeline for AI Agents

Build OCR pipelines for scanned PDFs with page routing, preprocessing, layout recovery, evidence, evaluation, and safe handoffs to downstream AI agents.

Trust and safety Aug 16, 2026

Accessible PDFs: convert to HTML, or remediate in place?

Convert a PDF to accessible HTML, or tag the PDF in place? Who argues each side, what the evidence says, and a sourced way to decide per document.

Trust and safety Jun 1, 2026

PDF accessibility checker: how to use a risk report

Use a PDF accessibility checker to triage risk: what to fix first, what needs human review, and what not to claim yet.

Agent engineering Aug 31, 2026

How to add a custom okraPDF connector in Claude Desktop

Connect okraPDF's MCP server to Claude Desktop in under a minute. Paste one URL, sign in, and start using PDF tools with no API key.

Trust and safety Aug 31, 2026

How to Redact Documents Securely: A Developer's Guide

How to redact documents so the data is actually gone: PII discovery, permanent removal, OCR and scanned files, metadata scrubbing, and an API workflow.

Agent engineering May 9, 2026

PDF to HTML: what 'HTML' should mean from a PDF API

Beyond raster screenshots: the four PDF-to-HTML artifacts okraPDF ships — hosted, reader, embeddable, configurable. With curl recipes.

Agent engineering May 6, 2026

OpenGraph links for PDF citations

How to turn page, target, and bbox evidence into shareable citation links with hover previews, GIF unfurls, and exact source pixels.

Extraction Jun 18, 2026

Convert a bank statement PDF to CSV (3 ways)

Three ways to turn a PDF bank statement into CSV — drag-drop tool, copy-by-hand, or a paid bookkeeping subscription. Step-by-step, ranked by speed and accuracy.

Extraction Jun 18, 2026

Stop writing regex for PDF extraction

Regex works until the PDF changes. Use schemas, source evidence, and validation when invoices, bank statements, or contracts become JSON.