# Changelog of PDF to Markdown & RAG Chunks Extractor (`mmdincer/pdf-rag-extractor`) Actor

- **URL**: https://apify.com/mmdincer/pdf-rag-extractor/changelog.md
- **Full Actor documentation**: https://apify.com/mmdincer/pdf-rag-extractor.md

## Changelog

### 1.0.12 — 2026-11-04

- Pages whose columns sit on the same baseline are written as Markdown tables, so a price list or a results table stays a table in the chunks. Prose and two-column body text are unchanged. No price or event-name change.

### 1.0.11 — 2026-11-04

- Documentation only: Related Actors now lists the audio transcriber and the video scene extractor as the RAG family. No code or pricing change.

### 1.0.0 — 2026-11-03

First release.

- One document row per PDF: full text, Markdown with headings and paragraphs, metadata, page / word / token counts.
- Token-counted chunks with page ranges, nearest heading and an overlap you control, counted with cl100k_base.
- Reads PDF links from another Actor's dataset (`datasetId`), so a crawl can be chunked in the same pipeline.
- Page window (`startPage`, `maxPages`) to skip covers and tables of contents.
- Scanned PDFs, failed downloads, password-protected and corrupt files write a row and are not charged.
- No browser, no OCR, no proxy by default.