Model card: unified repo + LiteRT/.tflite (drop ONNX; point to Desert-Ant-Labs/redact)

README.md

@@ -32,7 +32,9 @@ tags:
 - redaction
 - token-classification
 - on-device
-- onnx
+- litert
+- tflite
+- core-ml
 - multilingual
 pipeline_tag: token-classification
 ---
@@ -43,18 +45,20 @@ Detects and redacts personal data (names, addresses, emails, phone numbers,
 cards, IBANs, national IDs and more) in text across **all 24 official EU
 languages** (Latin, Greek and Cyrillic scripts). A BIOES token classifier plus a
 portable, dependency-free deterministic layer for structured IDs. The deployable
-model is **~13.7 MB** (int4 ONNX); with the tokenizer the total on-device
-footprint is **~16 MB**.
+model is **~11.6 MB** (4-bit Core ML on Apple) or **~24 MB** (int8 LiteRT
+`.tflite` on Android, Linux and the web).
 
 > `"Call Anna Kovács at anna@example.hu, IBAN GB29NWBK60161331926819"` →
 > `"Call [GIVEN_NAME] [SURNAME] at [EMAIL], IBAN [BANK_ACCOUNT]"`
 
 ## Try it
 
+All platforms ship from one repo: **[Desert-Ant-Labs/redact](https://github.com/Desert-Ant-Labs/redact)** (Swift, Kotlin, and JavaScript in a single codebase).
+
 - **Live demo:** [desert-ant-labs/redact-demo](https://huggingface.co/spaces/desert-ant-labs/redact-demo): paste text and watch PII get highlighted or masked, fully in your browser.
-- **iOS / macOS / tvOS / visionOS:** [`redact-swift`](https://github.com/Desert-Ant-Labs/redact-swift): the Swift SDK (Swift Package Manager) with a built-in demo app. It bundles the compiled Core ML model below.
-- **Node / browser (JavaScript / TypeScript):** [`redact-js`](https://github.com/Desert-Ant-Labs/redact-js) (`npm i @desert-ant-labs/redact`): fetches the model from this repo (transformers.js) and caches it.
-- **JVM / Android (Kotlin):** [`redact-kotlin`](https://github.com/Desert-Ant-Labs/redact-kotlin) (JitPack): ONNX Runtime with the bundled int4 model.
+- **iOS / macOS / tvOS / visionOS:** the Swift SDK (Swift Package Manager) with a built-in demo app. It bundles the compiled Core ML model below.
+- **Android / JVM (Kotlin):** Maven Central `ai.desertant:redact` — LiteRT (`.tflite`), with the model downloaded on demand or bundled via `ai.desertant:redact-tflite-resources`.
+- **Node / browser (JavaScript / TypeScript):** `npm i @desert-ant-labs/redact @litertjs/core` — a local WebAssembly pipeline with [LiteRT.js](https://www.npmjs.com/package/@litertjs/core) inference; fetches the model from this repo and caches it.
 
 ```swift
 import Redact
@@ -68,14 +72,18 @@ r.redactedText   // "Email [GIVEN_NAME_1] [SURNAME_1] at [EMAIL_1]."
 
 | File | Format | Size | Contents |
 |---|---|---:|---|
-| `redact.onnx` | ONNX (int4, opset 21) | ~13.7 MB | 4-bit-quantized model, batch=1, ready for on-device runtimes |
-| `redact.mlmodelc` | Compiled Core ML (4-bit) | ~11.6 MB | Palettized model, ready to load on Apple platforms (used by `redact-swift`) |
+| `redact.tflite` | LiteRT / TFLite (int8) | ~24 MB | Dynamic-range int8, fixed `[1, 256]` window; runs on Android, Linux and the web (used by the Kotlin and JavaScript SDKs) |
+| `redact.mlmodelc` | Compiled Core ML (4-bit) | ~11.6 MB | Palettized model, ready to load on Apple platforms (used by the Swift SDK) |
 | `redact.pt` | PyTorch checkpoint | ~90 MB | Full-precision weights + config (for retraining / other runtimes) |
 | `config.json` | JSON | tiny | Transformer + label config |
 | `tokenizer.json`, `tokenizer_config.json` | JSON | ~2.3 MB | EU-trimmed (31,475-piece) SentencePiece tokenizer (XLM-R lineage) |
+| `redact_tokenizer.bin` | Binary | ~0.4 MB | Compact SentencePiece vocab bundled by the SDKs |
 | `labels.json` | JSON | tiny | BIOES `id2label` / `label2id` |
 | `redact_meta.json` | JSON | tiny | Public labels, deterministic-owner labels, recommended thresholds, base-model info |
 
+Older revisions (tags `v0.2.x`) additionally carry `redact.onnx` for SDK
+versions that predate the LiteRT migration.
+
 ## Taxonomy (20 public labels)
 
 `GIVEN_NAME`, `SURNAME`, `STREET_NAME`, `BUILDING_NUMBER`, `SECONDARY_ADDRESS`,