Developer builds browser-only PDF OCR and signing tools with no server uploads

·1 views

A developer has released two browser-based PDF tools — one that makes scanned PDFs searchable via OCR, and another that adds signatures — neither of which uploads files to any server. Both tools run entirely within the browser tab, and all data is discarded once the tab is closed. The OCR tool uses pdf.js to render pages, tesseract.js for text recognition, and pdf-lib to embed an invisible text layer over the original scan. A key technical challenge was making WebAssembly-based OCR work under a strict Content-Security-Policy without relying on a CDN, which the developer solved by vendoring all required assets from the same origin. The CSP is generated per request via a Cloudflare Worker, scoping the necessary 'wasm-unsafe-eval' permission only to the OCR route rather than the entire site.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Three Practical Video Compression Workflows for Web Teams and Their Trade-offs

Web teams handling video assets often struggle less with encoding itself and more with choosing the right workflow for their size and structure. Desktop tools like HandBrake suit individual freelancers but create consistency, documentation, and audit problems for multi-person product teams. Engineering-led teams typically adopt a reproducible pipeline using FFmpeg wrapped in shell scripts or Makefiles, version-controlled via Git and run through CI for consistent output across machines. Marketing-led teams often prefer a spreadsheet-driven approach that lets non-engineers participate by tracking clips, resolutions, and target bitrates in a shared document. Each method carries distinct trade-offs around reproducibility, accessibility, and team ownership that should guide the final choice.

0 comments Read more at DEV Community

ProgrammingHacker News ·

Ancient Library launches with 1,060 Greek and Latin texts with word parsing

A new online platform called Ancient Library has been launched, offering access to 1,060 classical Greek and Latin texts. The site features an interactive tool that allows users to click on any word to receive an instant grammatical parse. The resource appears aimed at students, scholars, and enthusiasts of classical languages. The platform was shared on Hacker News, where it attracted early attention from the tech and academic community.

0 comments Read more at Hacker News

ProgrammingDEV Community ·

Why Python Remains the Top Recommended Language for Beginner Programmers

Python, created in the early 1990s by Guido van Rossum, has grown from a niche project into the world's most popular programming language. Educators, software engineers, and data scientists widely recommend it as the best starting point for those entering the technology field. The language is high-level, interpreted, and general-purpose, meaning it abstracts hardware complexity, executes code line by line for faster testing, and can be used across diverse applications. Python also supports multiple programming paradigms — procedural, object-oriented, and functional — allowing beginners to start simply and scale toward more complex architectures over time. Its readability and versatility make it a dominant choice across domains ranging from web development to advanced artificial intelligence.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Study of 170M Residential Proxy IPs Exposes Rapid Churn Undermining Fraud Detection

A 90-day analysis by IPinfo covering over 170 million residential proxy IP addresses, conducted between October and December 2025, found that this infrastructure changes far faster than most fraud detection systems assume. Residential proxy IPs were active for an average of just 4.56 days, with IPv6 addresses lasting only 1.29 days on average, and 60% of all observed IPs appearing only once. Around 78% of IPs vanished within 30 days of first being detected, making historical reputation data largely unreliable by the time it becomes actionable. The study also found that the same underlying IP infrastructure is frequently shared across multiple proxy providers simultaneously, further complicating provider-level identification. Together, rapid infrastructure turnover and cross-provider overlap pose a significant challenge to reputation-based and risk-scoring systems used in abuse and fraud detection.

0 comments Read more at DEV Community