clean(liturgy): relabel 'AI' witnesses with honest source labels

[anantham]

Summary

Sweep across three chant pages to fix the by: 'AI' witness label that the user previously flagged as junk. The 34 affected entries are all grounded translations; they were just mislabeled. This PR renames them per page to reflect the actual source:

File	Old label	Count	New label
morning-chants.ts	'AI'	19	'MAPLE chant text' — these are the simple English phrasings MAPLE actually chants ("I take refuge in the Buddha", etc.)
bodhicitta-dedication.ts	'AI'	4	'Standard English (Tibetan tradition)' — the canonical English rendering of the Tibetan dedication, chanted across lineages
enmei-jikku-kannon-gyo.ts	'AI'	11	'Literal English gloss' — character-by-character literal English from the Chinese kanji, kept as a comparison against the published Soto Zen / Red Cedar Zen translations

Also updated one curator note in enmei-jikku-kannon-gyo.ts that referenced the old 'AI' label.

EJKG note (no structural change needed)

While investigating, I confirmed both EJKG files (MAPLE and Bodhi) are already at full multi-script-witness depth — 10 chant-line segments each, Sino-Japanese + Hanzi + word-level glosses + per-kanji scriptMorphemes + per-segment notes. The MAPLE version has 3 witnesses (now: MAPLE chant text / Soto Zen / Red Cedar Zen); the Bodhi version has 1 (Bodhi Sangha). No structural work needed there.

Audit findings — what's left

Surveyed all 22 chant files (excluding helpers/index/sanghas/concepts):

Already at full depth (Heart Sutra benchmark):

• heart-sutra (MAPLE + Bodhi)
• sho-sai-myo-kichijo-darani
• shin-jin-no-mei, hokyo-zanmai, song-of-zazen
• enmei-jikku-kannon-gyo (MAPLE + Bodhi)
• morning-chants

Shallow (currently prose-commentary, 0 witnesses):

• precepts — fundamentally English commentary, no foreign-language source to add (Dhp 183 source is one Pali verse; the rest are Bodhidharma/Dōgen voicings in English)
• jade-method — MAPLE-composed English, no foreign source
• way-of-compassion — Pali source exists (Karaṇīya Metta Sutta = Sn 1.8); could be authored at depth (task #78)
• bodhisattva-vow-torei — Japanese (Hakuin lineage); could be authored at depth (task #79)
• dedication-and-evening-call — Japanese ceremonial; could be authored at depth (task #80)

Tasks #78-80 capture the next-pass authoring work. Precepts and Jade Method are intentionally left as-is — their content is the English commentary itself, not a foreign chant.

Test plan

• npx vitest run tests/components/liturgy/ — 729 pass, 57 skipped. No new test cases needed (mechanical label change).
• npx tsc --noEmit — no new errors (pre-existing song-of-zazen.ts error unchanged).
• Manual verification on Vercel preview: the MAPLE EJKG page should now show "Literal English gloss" as the first witness in the dots tooltip instead of "AI"; same for the other two pages.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
lexicon-forge	Ready	Preview, Comment	May 19, 2026 10:54am