pg_pageindex

PageIndex, but the work happens in Postgres.
Parse PDF/Markdown to structure JSON, run tree search and helpers—all from SQL.

_{Extension v1.0 · schema pageindex · upstream PageIndex}

---

Running in production? Read docs/install.md before you blame the extension for MuPDF path drama.¹

30-second start

CREATE EXTENSION pg_pageindex;

SELECT pageindex.build_pdf('/tmp/paper.pdf');
-- jsonb: document tree / structure for downstream SQL

flowchart LR
  A[(PDF / MD path)] --> B[pageindex.build_*]
  B --> C{{jsonb document}}
  C --> D[pageindex.tree_search]
  D --> E[(hits / subtree)]

Loading

Note

build_* reads files as the database server OS user, not your laptop login. Paths must exist on the server filesystem.

Documentation

Full write-ups live under docs/ (all lowercase filenames—docs/readme.md is the map):

docs/install.md	Build, make install, runtime MuPDF, CREATE EXTENSION
docs/api.md	Every pageindex.* function, grouped
docs/architecture.md	C ↔ Go bridge, .so layout, failure modes
docs/development.md	CI, make bridge-bump, local loop
docs/contributing.md	PR hygiene, changelog, PGXN
docs/security.md	Private disclosure, scope
docs/conduct.md	Code of Conduct pointer
docs/packaging.md	META.json, releases
docs/history.md	Link to CHANGELOG.md

Too lazy to click—give me the shell bits

git clone https://github.com/neurondb/pg_pageindex.git
cd pg_pageindex
make
sudo make install   # or: sudo make install PG_CONFIG=/wherever/pg_config

Then CREATE EXTENSION in psql. If the bridge cannot load MuPDF, the backend tends to die loudly—see docs/install.md.

Repo pointers

• License: LICENSE
• Changelog: CHANGELOG.md
• Authors: AUTHORS
• CoC: CODE_OF_CONDUCT.md
• Contributing (stub → full doc): CONTRIBUTING.md
• PGXN: META.json · docs/packaging.md

Footnotes

1. The bridge still needs MuPDF shared objects at runtime even though the Go side is built with -tags=nocgo. Details: docs/install.md. ↩