Phrasebook: multi-alias + French source keys, plus misses-script stub

Forbid parenthetical glosses in LLM replies

Stage 4: split translate/reply UI + CPU-safe TTS + reply-not-translate prompt

Per-turn telemetry + RAG few-shot on phrasebook miss

Remove stale Qwen references from minimal-baseline docstrings

Ground-zero Stages 1–3: dialect anchors + phrasebook short-circuit + Aya-Expanse

Fix jiwer crash on post-normalisation empty refs; register SLR106/105 datasets

Add Adlam/Pular Fula integration: transliterator + 3 new datasets + normalisation pipeline

Fix: add trust_remote_code=True for datasets 3.x compatibility

Phase 3: Voice-to-Voice S2S pipeline — F5-TTS, LLM brain, CER metric

Fix Self-Teaching tab: float sliders, deduplication, Kaggle API fallback

Fix Wikipedia and HF harvest buttons

Add mass training + internet self-teaching features

Add multi-user speaker profiles, collective voice, and mode toggle

Add confidence loop, curiosity engine, and lightweight TTS

Fix Bambara TTS red state + surface detailed errors in UI

Fix: install maliba-ai lazily at runtime, not at build time

Phase 2: Waxal TTS — Bambara voice output + Fula training notebook

Fix: switch LLM to Qwen2.5-72B-Instruct (Gemma not on HF free tier)

Phase 1: Sahel-Voice-Lab — The Memory Loop

Add phrase library + fuzzy matcher for general Bambara/Fula conversation

Improve Bambara/Fula quality + add full correction UI

Switch training dataset from google/waxal to google/fleurs

Initial commit: Sahel-Agri Voice AI