PDF to LlamaIndex JSON

How to Convert a PDF to LlamaIndex JSON

1. Set your options first. Chunk Size (characters) starts at 1024, and Chunk Overlap starts at 200. Include Metadata is Yes by default.
2. Add the PDF. Drop it onto the upload area or click Choose PDF File. Processing starts right away.
3. Check the counts. The page, character and chunk totals appear next to the JSON.
4. Read the LlamaIndex JSON Output box to make sure the text looks right.
5. Click Copy JSON or Download JSON. The download is named after your PDF with _llamaindex.json added.
6. Need a lighter format? Copy RAG Format gives a shorter structure with chunk text and page previews.

Why Use This PDF to LlamaIndex JSON Converter

• Your file stays on your device. The PDF is read and chunked by code running in your browser. Nothing is sent to a server.
• Chunking you control. Pick a chunk size from 256 to 4096 characters and an overlap. Chunks end at a sentence, line break or space near the limit instead of cutting a word in half.
• Everything in one file. The JSON has a document with the full text, a nodes list of chunks with start and end positions, and page nodes with one entry per page.
• Two copy formats. Take the full LlamaIndex-style JSON, or a smaller RAG format.

Good to know: the output is shaped like LlamaIndex Document and TextNode objects, but it is a plain JSON file, so you still load it with your own code. The file holds one document, its chunk nodes and per-page nodes, so use nodes if you only want chunks. Chunk size counts characters, not tokens. Changing a setting after you load a PDF redoes the file with the new value. Keep overlap well below chunk size. Scanned PDFs have no text to extract, and there is no OCR. For plain per-page JSON, see PDF to JSON.

FAQ

How do I convert a PDF to LlamaIndex JSON?

Set chunk size, overlap and metadata, then drop your PDF on the upload area. The JSON appears on the right, ready to copy or download.

What chunk size and overlap should I use?

A common starting point is 1024 characters with 200 overlap, which are the defaults. Smaller chunks suit short, precise answers. Larger ones keep more context. Keep the overlap well below the chunk size.

Can I load the file straight into LlamaIndex?

Not automatically. The fields follow LlamaIndex Document and TextNode objects, but you still read the JSON with your own script and build the objects. Use the nodes list for chunks.

What does Include Metadata add?

With Yes, the document metadata includes the file name, page count, total characters, extraction date and the parser name. With No, only the source file name is kept.

Does it work on scanned PDFs?

No. It reads text already stored in the PDF. A scan or photo gives empty text, because there is no OCR step. Run an OCR tool first.

Is my PDF uploaded anywhere, and is there a size limit?

No, the PDF is processed in your browser and never leaves your device. Your file size limit is shown on this page, and a free account raises it.