languages, carrier limits & safety rules
Every token classifier, privacy strip rule, and SMS/USSD packet boundary enforced offline.
| Code | Rule / Pack | Guarantee | Layer |
|---|---|---|---|
sw | Swahili (Kiswahili) | Token glossary across clinic triage, crops, pests, mobile money. | GSM-7 Ready |
lg | Luganda (Oluganda) | Supports Kampala & central Uganda code-switching + urgent terms. | GSM-7 Ready |
yo | Yoruba (Ede Yoruba) | Diacritic awareness (ẹ, ọ, ṣ) & automatic UCS-2 fallback. | UCS-2 Safe |
ha | Hausa | Sahel & Northern Nigeria health, farm, and marketplace terminology. | GSM-7 Ready |
fr | French (Francais) | Francophone Africa clinical, agritech & mobile balance terms. | GSM-7 / UCS-2 |
en / pcm / nyn | English, Pidgin & Runyankole | Global baseline + West African Pidgin + Western Uganda vocabulary. | 0 ms Lexicon |
PII-01 | Local PII Redaction | Masks PIN/OTP, phone numbers, and national IDs before cloud calls. | Local Scrub |
ESC-01 | Extension Worker Escalation | Flags urgency keywords and enqueues SMS receipt into SQLite WAL. | WAL Outbox |
GSM-7 / UCS-2 | SMS & USSD Packetizer | Enforces 182/160 GSM-7 vs 80/70 UCS-2 boundaries cleanly. | 182B USSD |
VAD-16K | Pure Python PCM Telemetry | Extracts RMS energy, SNR dB, and 16 kbps Opus data footprint. | Stdlib PCM |