Tools / Convert PDF for RAG
Convert PDF for RAG
Structured chunks for retrieval systems: heading paths, page spans, sentence-aware splits. A product feature — not our primary acquisition channel.
Drop a PDF here
or drag and drop a file
Advanced options
Limits: up to 100 pages · OCR up to 25 pages
- No account required
- Optional local OCR
- Files up to 40 MB
Converting in your browser
Preparing your PDF…
- Checking document
- Detecting page structure
- Reading scanned pages (OCR)
- Reading tables and images
- Cleaning text
- Creating Markdown
Conversion complete
View conversion details
Review recommended
View removed headers and footers
More exports
Nothing was uploaded. Clear the page to discard results from memory.
What you get
JSONL chunks with heading_path, page_start, page_end for retrieval pipelines.
- RAG chunks (.jsonl)
- Structured JSON
- Markdown (.md)
- ZIP package
Limits: RAG search volume is small; use this as an export profile after a strong Markdown conversion.
How it works
- Upload the PDF you want to index.
- We clean content and split by headings and sensible boundaries.
- Download chunks.jsonl for your vector store.
Example output
Whole document
Installation Manual — entire document text in one block spanning 48 pages...
RAG chunk (JSONL)
{"chunk_id": "c003", "heading_path": ["Installation", "Wiring"], "page_start": 12, "page_end": 13, "text": "Connect the ground wire before power."}
Use cases
- Populate RAG knowledge bases
- Prepare corpora for embeddings
Related tools
FAQ
What is in each chunk?
chunk_id, heading_path, page_start, page_end and text — with heading context attached.
Do you cut mid-sentence?
We avoid splitting mid-sentence and keep tables as whole units where possible.
Format of the download?
JSON Lines (.jsonl), one chunk object per line.
Is this the main product?
No. The main product is PDF to Markdown; RAG export is a secondary workflow.