LUGHA is the African community-powered linguistic infrastructure platform for African languages, collecting and preserving vocabulary, translations, pronunciations, stories, and cultural knowledge to power future language technologies and AI.
Lugha .
Verified African language infrastructure — a community-governed knowledge graph
that speakers teach and developers build on . Words, proverbs, pronunciation audio, confidence scores,
and licensed dataset exports for the next generation of African-language AI.
Not another chatbot — the datastore underneath one
Speakers verify. Communities score trust. Developers export and integrate.
Most AI has never learned African languages deeply. Lugha fixes the
data layer : structured terms, glosses, examples, heritage content, pronunciation recordings,
provenance, and CC-BY-4.0 licensing — exposed through a versioned HTTP API
and filtered corpus downloads.
---
## ✦ Two sides of the platform
Guardians
Teach & preserve
Contribute words, proverbs, phrases, and heritage stories. Record pronunciation.
Tag dialect, region, and source. Community votes shape confidence scores
— not a single authority, not a black-box model.
Developers
Build & train
Query the knowledge API live, or download filtered JSONL / CSV corpora.
Fetch pronunciation audio per entry. Use API keys with rate limits.
Plug Lugha into dictionaries, tutors, RAG chatbots, or native-language AI pipelines.
---
## ✦ What is in the MVP today
🎙
Teach with provenance
Words, proverbs, examples, dialect, region, source type, and license on every contribution.
Audio stored in MinIO (S3-compatible), streamed via /api/media .
📦
Dataset export
GET /api/v1/datasets/export — filter by language, confidence, content type,
dialect, license, has_audio . Download up to 5,000 rows as JSONL or CSV with a manifest line.
🔗
Knowledge API (v1)
Search entries with confidence scores, verification status, and audio URLs.
Community verify endpoint. Designed for live apps and RAG retrieval.
🔑
API keys & rate limits
Create keys from the Developers page. Authenticate with X-API-Key .
Separate bu …