[Omniwot]

Description

Docs-only change: adds a short "Readers maintained in their own repositories" list under
Data Loaders on the Integrations page, with one entry.

Section Anything is an Apache-2.0 library that
finds the sections of a PDF (headings, levels, pages, character spans). Its reader,
llama-index-readers-section-anything on PyPI, returns one Document per section with the
heading path as flat metadata and PARENT/CHILD/PREVIOUS/NEXT relationships, and works as a
SimpleDirectoryReader file extractor. Its default mode uses rules only (no model, no network).

pip install llama-index-readers-section-anything

from llama_index.core import SimpleDirectoryReader
from llama_index.readers.section_anything import SectionAnythingReader

reader = SectionAnythingReader()  # mode="free": rules only, no model, no cost
docs = reader.load_data("paper.pdf")
docs[2].metadata["heading_path"]  # '1 Introduction > 1.1 Scope'
# or for a folder: SimpleDirectoryReader("papers/", file_extractor={".pdf": reader}).load_data()

What it adds: one Document per real section of the PDF, with its heading path, pages and
character span, instead of one per page. In our Qasper test (200 papers, 707 questions),
section-bounded chunks retrieved the gold evidence more often than fixed-size chunks, mostly
because they are smaller, but answer quality did not change
(write-up).

Since new integration packages are maintained outside this repo (CONTRIBUTING.md), a place on
this page to link them seemed useful; happy to move or reword it.

AI use: the reader and this text were drafted with an AI assistant and reviewed by me; the
reader's tests run offline on a generated PDF.

New Package?

• Yes
• No

Version Bump?

• Yes
• No

Type of Change

• This change requires a documentation update

(docs-only change)

How Has This Been Tested?

• I added new unit tests to cover this change
• I believe this change is already covered by existing unit tests

(docs-only; the link and the package were checked by hand)

Suggested Checklist:

• I have performed a self-review of my own code
• I have made corresponding changes to the documentation
• My changes generate no new warnings