Lanna Manuscript Wiki

A small, zero-dependency (Python stdlib only) local wiki that browses the products of the Lanna manuscript crawler in ../crawler/. Double-click Wiki.command (or run python3 wiki.py) and it opens at http://127.0.0.1:4190.

It is a read-only view over the crawler's outputs — it never writes to the catalog:

• ../crawler/catalog.db — the SQLite catalog (sources, manuscripts, images, crawl_log) that the crawler upserts into.
• ../crawler/store/ — the content-addressed image store (store/<sha256[:2]>/<sha256>); the wiki serves page images from here by checksum.

Your own additions live in side files under wiki/data/, keyed so they survive re-crawls and DB rebuilds:

• data/annotations.json — your notes & tags, keyed by source::identifier.
• data/ocr.json — OCR text, keyed by image sha256 (so OCR follows the image bytes, not a row id).

What you can do

• Browse & search every catalogued manuscript. Facets come straight from the crawler's normalized tokens (crawler/genre.py): genre, script, material, language, province, source, century, and a Research priority filter (divination_omen, astrology, magic_ritual). Priority manuscripts are sorted first and flagged.
• Read a manuscript — every catalog column, the untouched raw_metadata blob (collapsible), and an image viewer paging through the stored page images with an Image / OCR text toggle. Links out to source_url and the IIIF manifest when present.
• OCR layer — OCR the stored images from the manuscript page; text is saved in data/ocr.json by image checksum.
• Notes & tags — per-manuscript, saved locally.
• Crawl status (/status) — counts, the sources table, and the recent crawl_log, so you can see harvest progress. Renders gracefully before the DB even exists.

Config

Environment overrides (all optional):

• CATALOG_DB — path to the catalog (default ../crawler/catalog.db).
• STORE_DIR — path to the image store (default ../crawler/store).
• PORT — default 4190.

OCR

From a manuscript page, pick a language and press Run OCR. It shells out to the Tesseract CLI (brew install tesseract tesseract-lang) over each stored image and writes the text into data/ocr.json.

Lanna (Tham) script has no off-the-shelf OCR model. Thai (tha) is only a starting point for Thai-script portions — swap in a better engine in run_ocr()/run_engine when you have one. If Tesseract or a language pack is missing, the run reports exactly what to install and changes nothing.

Files

File	Purpose
wiki.py	The whole server + UI (Python stdlib, no deps).
Wiki.command	Double-click launcher.
data/annotations.json	Your notes & tags.
data/ocr.json	OCR text by image checksum.

The catalog and image store are produced by the crawler in ../crawler/; this wiki only reads them.

Licence

Records, prose, pages and the compilation: CC BY 4.0, attributed to wichaa. Code: MIT. Attribution is the only condition — name the work and link back, and the rest is yours: copy it, adapt it, sell it, index it, train on it.

Manuscript images and their thumbnails are not covered by that: rights remain with each holding library, and the record's source link names it. Wikipedia text carried into the articles stays CC BY-SA 4.0, Wikidata is CC0, OpenStreetMap data is ODbL. Each layer, with its attribution, is in NOTICE.txt; the verbatim licence text is in LICENSE.

---

Contact: Nan · nan@motdang.net · Sponsor: Ko-fi · Patreon