docs: refresh model card (provenance, consumption) + ship engine config.json

- New Provenance section: adopt the per-promotion convention (commit
messages record train-repo SHA + MLflow run) and backfill v1-ljspeech
best-effort (run lyrical-goat-689, export-fix a537e03).
- Fix stale consumption steps: the workspace file is now sonora.tflite
under Reference/Models/ resolved via prosodia_models.json roles;
drop the bogus cargo invocation for apps/tuner/build.sh.
- Add v1-ljspeech/config.json (locked 178-symbol vocab, sample_rate
22050) so the e2e artifacts are self-describing — the Prosodia engine
reads this adjacent to the model file. bootstrap.sh now fetches it.
- YAML metrics: add wer (ASR WER 0.000 on the fixed float32 export).

train-repo: n/a (docs/config only) run: n/a

README.md

@@ -15,6 +15,7 @@ datasets:
 - ljspeech
 metrics:
 - loss
+- wer
 pipeline_tag: text-to-speech
 ---
 
@@ -64,6 +65,26 @@ The model is distributed in multiple formats for execution on different hardware
 
 ---
 
+## Provenance
+
+**Convention (adopted 2026-07-13):** every artifact promotion into this registry records, in its
+commit message, the training-repo state and run that produced it — `train-repo: <Sonora GitHub
+SHA>` and `run: <MLflow run name>` (plus the export toolchain when relevant). Commits before this
+date predate the convention.
+
+**v1-ljspeech (backfilled, best-effort):**
+* **Training:** MLflow run `lyrical-goat-689` on `ai-lab-0` (`sonora_training` container),
+  2026-07-10 → 11; 260 epochs, checkpoint Epoch 199 selected at the validation-loss plateau.
+  The exact training-repo SHA was not recorded at run time (pre-convention).
+* **Export fix & re-export:** training repo commit
+  [`a537e03`](https://github.com/Artificial-Humanity/Sonora/commit/a537e03) (channels-last
+  encoder LayerNorm rewrite), re-exported 2026-07-12; fidelity gate
+  `scripts/export_fidelity_referee.py` (ONNX↔TFLite cosine 1.0000, ASR WER 0.000).
+* **Split lane:** same Epoch-199 weights, converted per-module with `litert-torch` via the
+  [litert-samples](https://github.com/google-ai-edge/litert-samples) Matcha recipe.
+
+---
+
 ## Training Environment & Hardware
 
 The model was trained locally on the `ai-lab-0` workstation using a containerized AMD ROCm configuration:
@@ -106,12 +127,12 @@ The model was trained locally on the `ai-lab-0` workstation using a containerize
 The compiled model weights are loaded dynamically by the `StageCoordinator` through the TFLite runtime C-API. 
 
 To use this model locally:
-1. Download `checkpoint_epoch=199_e2e_float32.tflite` (the verified, `f32`-I/O artifact; the float16 build needs a delegate / engine `f16` support — see the format notes above).
-2. Rename the file to `styletts2_lite.tflite` (or update your environment configs to load the custom filename).
-3. Place it in your shared `Models/` directory.
-4. Run your application:
+1. Download `checkpoint_epoch=199_e2e_float32.tflite` (the verified, `f32`-I/O artifact; the float16 build needs a delegate / engine `f16` support — see the format notes above) and the adjacent `v1-ljspeech/config.json` (the engine contract: locked 178-symbol vocab + `sample_rate` — the Prosodia engine reads it next to the model file).
+2. Place both next to each other where your config points; name the model file whatever your role config declares (the Artificial-Humanity workspace uses `sonora.tflite`, resolved via the `actor` role in Prosodia's `prosodia_models.json`).
+3. In the Artificial-Humanity workspace this is automated: the umbrella `bootstrap.sh` copies the blessed artifact + config into `Reference/Models/`.
+4. Build and run the harness:
    ```bash
-   cargo run --package tuner-app
+   Prosodia/apps/tuner/build.sh
    ```
 
 ### Input/Output Tensor Node Contract

v1-ljspeech/config.json

@@ -0,0 +1,183 @@
+{
+  "sample_rate": 22050,
+  "vocab": {
+    "_": 0,
+    ";": 1,
+    ":": 2,
+    ",": 3,
+    ".": 4,
+    "!": 5,
+    "?": 6,
+    "¡": 7,
+    "¿": 8,
+    "—": 9,
+    "…": 10,
+    "\"": 11,
+    "«": 12,
+    "»": 13,
+    "“": 14,
+    "”": 15,
+    " ": 16,
+    "A": 17,
+    "B": 18,
+    "C": 19,
+    "D": 20,
+    "E": 21,
+    "F": 22,
+    "G": 23,
+    "H": 24,
+    "I": 25,
+    "J": 26,
+    "K": 27,
+    "L": 28,
+    "M": 29,
+    "N": 30,
+    "O": 31,
+    "P": 32,
+    "Q": 33,
+    "R": 34,
+    "S": 35,
+    "T": 36,
+    "U": 37,
+    "V": 38,
+    "W": 39,
+    "X": 40,
+    "Y": 41,
+    "Z": 42,
+    "a": 43,
+    "b": 44,
+    "c": 45,
+    "d": 46,
+    "e": 47,
+    "f": 48,
+    "g": 49,
+    "h": 50,
+    "i": 51,
+    "j": 52,
+    "k": 53,
+    "l": 54,
+    "m": 55,
+    "n": 56,
+    "o": 57,
+    "p": 58,
+    "q": 59,
+    "r": 60,
+    "s": 61,
+    "t": 62,
+    "u": 63,
+    "v": 64,
+    "w": 65,
+    "x": 66,
+    "y": 67,
+    "z": 68,
+    "ɑ": 69,
+    "ɐ": 70,
+    "ɒ": 71,
+    "æ": 72,
+    "ɓ": 73,
+    "ʙ": 74,
+    "β": 75,
+    "ɔ": 76,
+    "ɕ": 77,
+    "ç": 78,
+    "ɗ": 79,
+    "ɖ": 80,
+    "ð": 81,
+    "ʤ": 82,
+    "ə": 83,
+    "ɘ": 84,
+    "ɚ": 85,
+    "ɛ": 86,
+    "ɜ": 87,
+    "ɝ": 88,
+    "ɞ": 89,
+    "ɟ": 90,
+    "ʄ": 91,
+    "ɡ": 92,
+    "ɠ": 93,
+    "ɢ": 94,
+    "ʛ": 95,
+    "ɦ": 96,
+    "ɧ": 97,
+    "ħ": 98,
+    "ɥ": 99,
+    "ʜ": 100,
+    "ɨ": 101,
+    "ɪ": 102,
+    "ʝ": 103,
+    "ɭ": 104,
+    "ɬ": 105,
+    "ɫ": 106,
+    "ɮ": 107,
+    "ʟ": 108,
+    "ɱ": 109,
+    "ɯ": 110,
+    "ɰ": 111,
+    "ŋ": 112,
+    "ɳ": 113,
+    "ɲ": 114,
+    "ɴ": 115,
+    "ø": 116,
+    "ɵ": 117,
+    "ɸ": 118,
+    "θ": 119,
+    "œ": 120,
+    "ɶ": 121,
+    "ʘ": 122,
+    "ɹ": 123,
+    "ɺ": 124,
+    "ɾ": 125,
+    "ɻ": 126,
+    "ʀ": 127,
+    "ʁ": 128,
+    "ɽ": 129,
+    "ʂ": 130,
+    "ʃ": 131,
+    "ʈ": 132,
+    "ʧ": 133,
+    "ʉ": 134,
+    "ʊ": 135,
+    "ʋ": 136,
+    "ⱱ": 137,
+    "ʌ": 138,
+    "ɣ": 139,
+    "ɤ": 140,
+    "ʍ": 141,
+    "χ": 142,
+    "ʎ": 143,
+    "ʏ": 144,
+    "ʑ": 145,
+    "ʐ": 146,
+    "ʒ": 147,
+    "ʔ": 148,
+    "ʡ": 149,
+    "ʕ": 150,
+    "ʢ": 151,
+    "ǀ": 152,
+    "ǁ": 153,
+    "ǂ": 154,
+    "ǃ": 155,
+    "ˈ": 156,
+    "ˌ": 157,
+    "ː": 158,
+    "ˑ": 159,
+    "ʼ": 160,
+    "ʴ": 161,
+    "ʰ": 162,
+    "ʱ": 163,
+    "ʲ": 164,
+    "ʷ": 165,
+    "ˠ": 166,
+    "ˤ": 167,
+    "˞": 168,
+    "↓": 169,
+    "↑": 170,
+    "→": 171,
+    "↗": 172,
+    "↘": 173,
+    "'": 174,
+    "̩": 175,
+    "ᵻ": 176,
+    "ᵊ": 177
+  }
+}