# termux-stt - Full Architecture & API Specification

## 1. Overview
termux-stt is an on-device Speech-to-Text (STT) and Speaker Diarization framework engineered specifically for the Android Termux environment.

## 2. Supported Engines
- whisper.cpp: High accuracy GGML-quantized STT
- vosk: Ultra-lightweight Kaldi-based STT & 128d X-Vector speaker embedding
- sherpa-onnx: Next-gen Kaldi ONNX runtime for streaming & offline STT
- hybrid: Vosk X-Vector + Whisper STT + Pure Python K-Means

## 3. Pure Python Math
Clustering and similarity calculation are implemented in pure Python to eliminate large native ML dependencies (numpy, scipy, scikit-learn).