Document Processing · head-to-head

liteparse vs docling

Pick liteparse when you need one figure, table or passage out of a document without loading the whole thing into context — it is fast, local, model-free, and needs no API key. Pick docling when the document itself has to become structured data: ten-plus input formats, OCR on scanned pages, and chunking a corpus for retrieval. Both are MIT and both run on your machine, so cost is not the deciding factor; the size of the job is.

Pick liteparse

Targeted extraction from a large document — look up a number, find a table, answer a question about a file — with minimal context spend.

Pick docling

Whole-document conversion, scanned or image-based pages, or preparing a document set for RAG.

Side by side

How they differ

Job shape

liteparse
Search and extract specific values
docling
Convert entire documents to structured output

Scanned pages / OCR

liteparse
No — its own entry points at the hosted llamaparse skill for that
docling
Yes

Formats

liteparse
PDF, DOCX, PPTX, XLSX, images
docling
PDF, DOCX, PPTX, XLSX, HTML, Markdown, AsciiDoc, CSV, images, audio, XML

RAG chunking

liteparse
No
docling
Built in

Context cost

liteparse
Low by design — searches rather than loading
docling
Higher — converts the whole document

Setup

liteparse
2 min, npx skills add, Node 18+, no API key
docling
5 min, pip install docling

Provenance

liteparse
Verified (LlamaIndex)
docling
Verified (Docling project, LF AI & Data)

Open liteparse or docling for the composite score, license, and exact install commands, or browse the full Document Processing category.

Verdict

The bottom line

Reach for liteparse first on any question that sounds like "what does this document say about X" — it is the cheaper answer and the one that will not blow your context window on a 200-page PDF. Move to docling when the answer is "all of it, structured", or when the pages are scans. They stack well: liteparse for interactive lookups during a task, docling for the batch job that builds an index.

FAQ

Common questions

Does liteparse need an API key?

No, and that is the point of choosing it over its sibling. It runs locally through the lit CLI with Node 18+. The llamaparse skill in the same LlamaIndex repo is the hosted, key-requiring option for harder layouts.

Which one should I use for a scanned contract?

docling. liteparse is model-free local extraction and its own catalogue entry says scanned-image OCR is not its job — it points at llamaparse instead. docling handles scanned PDFs directly.

Can I install both?

Yes, and it is a reasonable setup. Their descriptions trigger on different asks — a targeted lookup versus a conversion or chunking job — so the agent picks correctly without you routing manually.

Related

More head-to-head comparisons

Browse all best-skill guides or the full directory.