Equip: an open-source LMS for Bible schools — bilingual, scripture-aware, 4 weeks of real users

TL;DR. Equip is a free, open-source LMS built for Bible schools and small ministries. MIT-licensed. Bilingual Russian↔English out of the box — the same course renders in either language depending on who's reading. Self-hostable end-to-end on free tiers. 4 weeks in production: 14 real users, 3 published courses, 3 issued certificates. This post is what's in it, what's wired in, and how it's built.

. Live at

The Cyrillic side isn't a different file or a different route. It's the same React tree with useTranslation(), plus the scripture passage in the left rail swapped from KJV (1769) to Synodal (1876) by a server-side substitution layer — see next section.

The scripture-preservation constraint

Most translation APIs (Gemini, GPT-4, DeepL) will happily paraphrase a Bible verse if you feed it into them as part of a longer paragraph. For a course where the exact wording of John 3:16 matters, that's a non-starter. "For God so loved the world" and "Because God loved the world this much" are not interchangeable in a Bible-study context, no matter how close the meaning.

Equip solves this outside the LLM:

1. Pre-process incoming HTML, detect <blockquote> + scripture-reference pairs.


2. Swap the verse text for an opaque ASCII placeholder (VERSE_<hex>).


3. Send the prose-with-placeholders to Gemini Flash Lite with a system prompt that lists "preserve placeholders verbatim" as a hard rule.


4. After the model returns, restore each placeholder with the canonical text from bundled translations: KJV 1769 for English, Synodal 1876 for Russian.

The model never sees the verse text. The translation never paraphrases scripture. Implementation lives in app/services/bible/substitution.py and is about 200 lines.

There's a fallback in the system prompt for paraphrased quotes that the substitution layer can't confidently match (similarity < 0.80 to canonical) — "leave the original verse text untouched." That covers content where the substitution layer can't reliably do its job.

What's wired in

Equip is small in scope but talks to a few external services where it makes sense to. Each one is a deliberate choice, not a default:

The pattern across all of these: pick a single best-fit service per layer, stay on free tiers as long as possible, never hide the integration behind a paywall.

What it does (the standard LMS surface)

Courses → modules → chapters, rich content via TipTap (callouts, audio, embedded YouTube, images), quizzes (multiple choice, true/false, short-answer, essay), assignments with a teacher grading queue, certificates with teacher approval, gradebook, cohorts, calendar, announcements.

The architecture, briefly

Total infra cost at current usage: $0. Vercel, Supabase, Datadog, and Resend free tiers carry the whole thing.

Mobile, too

. Free. Sign up, you're in. Works for independent teachers, single-classroom contexts, students looking for Bible courses to enroll in. No credit card.

Self-host your own instance. Clone , in-platform certificate verification via signed JSON-LD, and a contribution path for translators to refine AI-translated content where wording matters most.

What I want to hear back

Especially from three groups:

• 
  Anyone running a small Bible school or ministry training program: what's blocking you from moving off your current setup? Particularly curious about Russian-speaking schools in CIS countries where the Synodal canonical-text handling matters.


• 
  Anyone who has built or contributed to a niche-domain LMS: what's the one constraint you wish you'd designed for from day one? I'd rather steal your lessons than learn them again.


• 
  Anyone considering forking Equip for a different niche (Sunday-school curriculum, a different religious tradition, secular adult-ed): tell me what'd need to change. Easier to plan the abstractions now than retrofit them later.