Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,51 @@
+---
+license: cc-by-4.0
+task_categories:
+  - automatic-speech-recognition
+language:
+  - en
+size_categories:
+  - n<1K
+tags:
+  - calibration
+  - librispeech
+---
+
+# STT Calibration Dataset
+
+Tiny calibration dataset for PersonalAssistant STT service. Used on first run to auto-tune speculative pre-transcription parameters.
+
+## Contents
+
+| File | Duration | Size | Purpose |
+|------|----------|------|---------|
+| `short.wav` | 3.5s | 110KB | RTF measurement + VAD onset latency |
+| `long.wav` | 23.3s | 729KB | Split quality calibration (whole vs split comparison) |
+| `very_long.wav` | 56.8s | 1.8MB | Multi-split calibration (find minimum safe split interval) |
+| `manifest.json` | - | 2KB | Sample metadata + reference transcriptions |
+
+Total: ~2.6MB
+
+## Source
+
+All audio from [LibriSpeech](https://www.openslr.org/12/) test-clean (CC BY 4.0). `very_long.wav` is 4 samples concatenated with 0.5s silence gaps.
+
+## Usage
+
+Downloaded automatically on first STT run via `huggingface_hub.snapshot_download`:
+
+```python
+from huggingface_hub import snapshot_download
+path = snapshot_download("cvxhull/stt-calibration")
+```
+
+Cached in `~/.cache/huggingface/hub/`. No re-download on subsequent runs.
+
+## Calibration Parameters
+
+| Parameter | How it's calibrated |
+|-----------|-------------------|
+| ASR RTF | Transcribe `short.wav`, measure wall time / audio duration |
+| Split interval | Derived from RTF: `clamp(target_latency / rtf, min_safe_split, buffer_timeout)` |
+| Min safe split | Transcribe `long.wav` whole vs split at [5s, 8s, 12s], find minimum with similarity >= 0.95 |
+| VAD onset latency | Feed `short.wav` through VAD, measure chunks until first detection |