PDF to LLM Markdown — RAG Chunks & OCR-Ready

Store: https://apify.com/plugiverse/pdf-to-llm-markdown
Actor id: o8HXWdpdrS10NmQMW · username plugiverse · LIMITED_PERMISSIONS · pay-per-event

Turn PDF URLs into clean, LLM-ready Markdown with optional RAG chunks and page markers. Text-layer pages are cheap; OCR is charged only for pages that need it. No HTML crawling, no DOCX/XLSX.

🤖 Agents & Apify MCP (start here)

Agents discover this Actor through Apify MCP (https://mcp.apify.com) — search tools for pdf llm markdown, pdf to markdown, RAG PDF chunks, or OCR PDF markdown.

Minimal call:

{
  "pdfUrls": ["https://raw.githubusercontent.com/mozilla/pdf.js/ba2edeae/web/compressed.tracemonkey-pldi-09.pdf"],
  "enableOcr": "never",
  "maxPages": 3,
  "chunkSize": 2000,
  "includePageMarkers": true
}

What you get back (one dataset row per URL):

• status: "ok" or "error"
• markdown: full text with <!-- page: N --> markers
• chunks[]: { index, text, pageStart, pageEnd, charCount } for RAG citations
• billing.eventsCharged: what was billed

Failed PDFs are free (status: "error", machine-readable error.code). A run never “succeeds while empty” when failOnAllErrors is true (default).

Eligible for agentic / PPE agent runs: LIMITED_PERMISSIONS + pay-per-event (no usage addon, no Standby).

What it does / when to use it

Use it when an agent or RAG pipeline already has PDF links (filings, manuals, papers) and needs Markdown plus citation-friendly chunks.
Don’t use it to crawl websites, convert Office docs, or unlock paywalled / password-protected PDFs (v1).

Input

Field	Type	Default	Notes
pdfUrls	string[]	required	http(s) PDF URLs; up to 100; duplicates removed
maxPages	int 1–500	100	Cap pages processed per PDF
chunkSize	int	2000	RAG chunk target size in characters; 0 disables chunks
enableOcr	never | auto | always	auto	OCR empty text-layer pages. OCR still stubbed → OCR_NOT_IMPLEMENTED on scans; text PDFs work.
includePageMarkers	bool	true	Insert <!-- page: N --> before each page
maxPdfMb	int 1–100	50	Download size cap
failOnAllErrors	bool	true	Run ends FAILED if every PDF fails

Output (one row per URL)

{
  "sourceUrl": "https://example.com/report.pdf",
  "status": "ok",
  "error": null,
  "pageCount": 12,
  "pagesProcessed": 12,
  "markdown": "<!-- page: 1 -->\n# Report\n…",
  "chunks": [{"index": 0, "text": "…", "pageStart": 1, "pageEnd": 2, "charCount": 1840}],
  "ocrUsed": false,
  "billing": {"eventsCharged": [{"event": "pdf-start", "count": 1}, {"event": "text-page", "count": 12}]},
  "fetchedAt": "2026-11-06T07:00:00Z"
}

Error codes (all free)

INPUT_INVALID, DOWNLOAD_FAILED, NOT_PDF, EMPTY_PDF, OCR_NOT_IMPLEMENTED, PDF_CORRUPT, TIMEOUT, TOO_LARGE, INTERNAL_ERROR (retryable), CHARGE_LIMIT_REACHED.

Pricing (pay per event)

Event	Price	When
apify-actor-start	$0.00005	Synthetic start (Apify covers first 5s compute)
pdf-start	$0.005	PDF converted successfully (status: ok)
text-page	$0.0002	Per page with a usable text layer
ocr-page	$0.008	Per page that required OCR

Error rows: $0. Example: a 20-page text PDF ≈ $0.005 + 20×$0.0002 = ~$0.009.

Scope lock (v1)

• PDF URLs only (http/https). Magic-byte %PDF check after download.
• PyMuPDF text-layer extraction.
• OCR stubbed until wired; use enableOcr: "never" for text-layer PDFs.
• No website crawl, no DOCX/XLSX, no password unlock beyond empty-password try.

Kill rule

Kill / unpublish if <20 external users or <$20 revenue by day 45 after Store publish.

Changelog

• 0.1.3: Agent README example matches input schema (chunkSize); MCP search keywords expanded.
• 0.1.2: Store public + output schema; agent README; AUTOMATION category; apify-actor-start.
• 0.1.0: text-layer convert, RAG chunks, PPE events, OCR stubbed.