Mnemosyne — Language Learning from Any Text

How it works

Paste

Any text — article, book, lyrics, conversation, legal document

Parse

Sentences split into learnable objects: vocabulary, conjugation, grammar, idiom, nuance

Learn

Per-word micro-lessons with morphological analysis and grammatical context

Review

FSRS scheduling surfaces the right words at the right moment

Progress

i+1 recommendations find sentences just beyond your current level

Languages

Full morphological analysis — full · Rich morphology via Stanza — rich · Vocabulary / dictionary mode — standard

Spanishfull

Frenchfull

Germanfull

Russianfull

Japanesefull

Portuguesefull

Italianfull

Englishfull

Finnishrich

Latinmorph

Koine Greekmorph

Koreanmorph

Hindimorph

Turkishmorph

Arabicvocab · RTL

Hebrewvocab · RTL

Mandarinvocab · CJK

API

Self-hosted FastAPI backend. Parse text, retrieve lessons, submit reviews, get recommendations.

POST /ingest
{
  "text": "Hola. Yo hablo español.",
  "language": "es",
  "content_type": "article"
}

→ sentences with learnable objects (vocabulary, conjugation, grammar, idiom, nuance)
→ deterministic UUID per object: same word always same ID
→ source_document_id for reading-progression tracking

POST /review  { "object_id": "...", "quality": 3 }
→ next_interval_days  (FSRS scheduling)

GET  /recommend?language=es
→ sentences at i+1 difficulty from your parse history

Self-host in 3 commands

cp .env.example .env
make build
make up
# API on :8000 · frontend: python -m http.server 8080 -d frontend

View on GitHub Architecture Write a plugin