Deduplicate a lead list in Python: five ways drop_duplicates misses
Missing emails merged into one, case and spaces, 1042 vs "1042", the emptier row kept, and one website spelled five ways. Each one measured in pandas 3.1
Latest articles
Read RSS, Atom and JSON Feed in Python: four feedparser traps
JSON Feed it can't read, strings it fetches, dates it invents and authors it drops, plus how to find a site's feed. Each one measured
Download files from a list of URLs in Python without getting burned
A size cap that works before and during the body, a content-type check, safe file names, and no private addresses. Each one measured against a local server
PDF to Markdown in Python: rebuilding headings, lists and paragraphs from font sizes
pdfminer.six gives you text boxes and font sizes; here are the seven rules we use to turn them into clean Markdown
OCR a scanned PDF in Python: why 300 dpi, and how to catch silent misreads
A scanned PDF is pictures of text. We rendered one page at 72, 150 and 300 dpi, read it with Tesseract, and measured what each resolution gets wrong
Offline OCR in a Manifest V3 Chrome extension with tesseract.js: five things we had to measure
Where the engine can run, mapping a drag onto a screenshot, why we enlarge 2x, dropping corner junk, and testing with the network cut
Transcribe audio and video files to text and SRT subtitles with about 50 lines of Python
A runnable example that turns a list of media links into .txt transcripts and .srt subtitles, with a cost cap
