fr-spell 1.0.8 → 1.0.9

package/README.cn.md

@@ -16,7 +16,7 @@ FR-SPELL 是一个用于法语词形还原与派生形态生成的 npm 包。
 
 免费社区版： [fr-spell](https://www.npmjs.com/package/fr-spell)
 
-| | [fr-spell (community)](https://www.npmjs.com/package/fr-spell) | @cndv3996/fr-spell-mini |
+| | [fr-spell (community)](https://www.npmjs.com/package/fr-spell) | fr-spell-mini |
 |---|---|---|
 | **词元模型大小** | 1.48 MB | 0.96 MB |
 | **派生模型大小** | 1.40 MB | 0.91 MB |
@@ -66,6 +66,21 @@ console.log(verb);
 { lemma: 'manger', wordType: 'VERB', person: 'FST_PL', mode: 'INDI', tense: 'PRES', output: 'mangeons', confidence: 0.9999864523, timeMs: 4.79 }
 ```
 
+
+## 浏览器用法
+
+```html
+<script src="./frspell.browser.js"></script>
+<script>
+	(async () => {
+		const predictor = await window.FrSpell({
+			modelBasePath: './models/community'
+		});
+		const result = await predictor.lemma('mangeons');
+		console.log(result);
+	})();
+</script>
+```
 ## 预测参数说明
 
 词元预测（lemma）：
@@ -118,79 +133,8 @@ console.log(verb);
 - 参考定义文件中还有更多时态名称，但本包实现目前只支持 `PRES`、`IMPA`、`FUTU`、`PASS`。
 - 对于名词/形容词派生，用户输入时不需要 `mode` 与 `tense`。
 
-## 运行测试
-
-```bash
-npm test
-```
-
-该命令会执行 test/test.js，并输出示例预测结果。
-
-## 查看帮助
-
-```bash
-npm run help
-```
-
-该命令会输出参数速查说明，覆盖 `lemma`、`nounDerive`、`adjeDerive`、`verbDerive`、`derive` 及 person/mode/tense 可用值。
-
-## 构建浏览器 Bundle
-
-```bash
-npm run build
-```
-
-构建产物：
-
-- `dist/frspell.browser.js`（单个 JS 文件，挂载 `window.FrSpell`）
-- `dist/models/community/*.onnx|*.json`（模型与词表等必需资源）
-
-构建完成后，将整个 `dist` 文件夹复制到前端项目中，然后使用：
-
-```html
-<script src="./dist/frspell.browser.js"></script>
-<script>
-	(async () => {
-		const predictor = await window.FrSpell({
-			modelBasePath: './dist/models/community'
-		});
-		const result = await predictor.lemma('mangeons');
-		console.log(result);
-	})();
-</script>
-```
-
-## 运行基准测试
-
-1) 先生成检查清单 JSON（每类 100 条）：
-
-```bash
-npm run benchmark:prepare
-```
-
-2) 运行全部基准测试：
-
-```bash
-npm run benchmark
-```
-
-3) 可选：按单项运行：
-
-```bash
-npm run benchmark:lemma
-npm run benchmark:noun
-npm run benchmark:verb
-npm run benchmark:adje
-```
-
 ## 基准结果（最近一次本地运行）
 
-基准命令：
-
-```bash
-npm run benchmark
-```
-
 结果：
 
 - 由变位预测词元：97/100，准确率 97.00%，平均 21.97 ms

package/README.fr.md

@@ -16,7 +16,7 @@ Le package s'appuie sur ONNX Runtime et des modèles INT8 quantifiés pour offri
 
 Une version communautaire gratuite est aussi disponible : [fr-spell](https://www.npmjs.com/package/fr-spell)
 
-| | [fr-spell (community)](https://www.npmjs.com/package/fr-spell) | @cndv3996/fr-spell-mini |
+| | [fr-spell (community)](https://www.npmjs.com/package/fr-spell) | fr-spell-mini |
 |---|---|---|
 | **Taille du modele de lemme** | 1.48 MB | 0.96 MB |
 | **Taille du modele de derive** | 1.40 MB | 0.91 MB |
@@ -66,6 +66,21 @@ Exemple de sortie a l'execution :
 { lemma: 'manger', wordType: 'VERB', person: 'FST_PL', mode: 'INDI', tense: 'PRES', output: 'mangeons', confidence: 0.9999864523, timeMs: 4.79 }
 ```
 
+
+## Utilisation dans le navigateur
+
+```html
+<script src="./frspell.browser.js"></script>
+<script>
+	(async () => {
+		const predictor = await window.FrSpell({
+			modelBasePath: './models/community'
+		});
+		const result = await predictor.lemma('mangeons');
+		console.log(result);
+	})();
+</script>
+```
 ## Parametres de prediction
 
 Prediction de lemme :
@@ -118,79 +133,8 @@ Note :
 - Le fichier de definitions d'origine contient plus de noms de temps, mais ce package prend actuellement en charge uniquement `PRES`, `IMPA`, `FUTU`, `PASS`.
 - Pour les appels nom/adjectif, `mode` et `tense` ne sont pas requis dans l'entree utilisateur.
 
-## Exécuter les tests
-
-```bash
-npm test
-```
-
-Cette commande exécute test/test.js et affiche des exemples de prédiction.
-
-## Afficher l'aide
-
-```bash
-npm run help
-```
-
-Cette commande affiche un guide rapide des paramètres pour `lemma`, `nounDerive`, `adjeDerive`, `verbDerive` et `derive`, avec les valeurs autorisées de person/mode/tense.
-
-## Construire le bundle navigateur
-
-```bash
-npm run build
-```
-
-Sortie de build :
-
-- `dist/frspell.browser.js` (fichier JS unique, expose `window.FrSpell`)
-- `dist/models/community/*.onnx|*.json` (ressources modèle et vocabulaire requises)
-
-Après le build, copiez tout le dossier `dist` dans votre projet frontend, puis utilisez :
-
-```html
-<script src="./dist/frspell.browser.js"></script>
-<script>
-	(async () => {
-		const predictor = await window.FrSpell({
-			modelBasePath: './dist/models/community'
-		});
-		const result = await predictor.lemma('mangeons');
-		console.log(result);
-	})();
-</script>
-```
-
-## Exécuter les benchmarks
-
-1) Générer les fichiers JSON de checklist (100 éléments chacun) :
-
-```bash
-npm run benchmark:prepare
-```
-
-2) Exécuter toutes les suites de benchmark :
-
-```bash
-npm run benchmark
-```
-
-3) Optionnel : exécuter une suite spécifique :
-
-```bash
-npm run benchmark:lemma
-npm run benchmark:noun
-npm run benchmark:verb
-npm run benchmark:adje
-```
-
 ## Résultats de benchmark (dernier run local)
 
-Commande de benchmark :
-
-```bash
-npm run benchmark
-```
-
 Résultats :
 
 - lemme depuis une conjugaison : 97/100, précision 97.00 %, moyenne 21.97 ms

package/README.md

@@ -16,7 +16,7 @@ The package runs with ONNX Runtime and quantized INT8 models for high speed and
 
 A free community version is also available: [fr-spell](https://www.npmjs.com/package/fr-spell)
 
-| | [fr-spell (community)](https://www.npmjs.com/package/fr-spell) | @cndv3996/fr-spell-mini |
+| | [fr-spell (community)](https://www.npmjs.com/package/fr-spell) | fr-spell-mini |
 |---|---|---|
 | **Lemma model size** | 1.48 MB | 0.96 MB |
 | **Derive model size** | 1.40 MB | 0.91 MB |
@@ -66,6 +66,21 @@ Sample runtime output:
 { lemma: 'manger', wordType: 'VERB', person: 'FST_PL', mode: 'INDI', tense: 'PRES', output: 'mangeons', confidence: 0.9999864523, timeMs: 4.79 }
 ```
 
+
+## Browser Usage
+
+```html
+<script src="./frspell.browser.js"></script>
+<script>
+	(async () => {
+		const predictor = await window.FrSpell({
+			modelBasePath: './models/community'
+		});
+		const result = await predictor.lemma('mangeons');
+		console.log(result);
+	})();
+</script>
+```
 ## Prediction Parameters
 
 Lemma prediction:
@@ -118,79 +133,8 @@ Note:
 - The original grammar definition file includes more tense names, but this package implementation currently supports only `PRES`, `IMPA`, `FUTU`, `PASS`.
 - For noun/adjective derive calls, `mode` and `tense` are not required in user input.
 
-## Run Test
-
-```bash
-npm test
-```
-
-This executes test/test.js and prints sample prediction outputs.
-
-## Run Help
-
-```bash
-npm run help
-```
-
-This prints a quick parameter reference for `lemma`, `nounDerive`, `adjeDerive`, `verbDerive`, and `derive`, including allowed person/mode/tense values.
-
-## Build Browser Bundle
-
-```bash
-npm run build
-```
-
-Build output:
-
-- `dist/frspell.browser.js` (single JS bundle, attach `window.FrSpell`)
-- `dist/models/community/*.onnx|*.json` (required model and vocab assets)
-
-After build, copy the whole `dist` folder into your frontend project, then use:
-
-```html
-<script src="./dist/frspell.browser.js"></script>
-<script>
-	(async () => {
-		const predictor = await window.FrSpell({
-			modelBasePath: './dist/models/community'
-		});
-		const result = await predictor.lemma('mangeons');
-		console.log(result);
-	})();
-</script>
-```
-
-## Run Benchmark
-
-1) Prepare checklist JSON files (100 items each):
-
-```bash
-npm run benchmark:prepare
-```
-
-2) Run all benchmark suites:
-
-```bash
-npm run benchmark
-```
-
-3) Optional single-suite runs:
-
-```bash
-npm run benchmark:lemma
-npm run benchmark:noun
-npm run benchmark:verb
-npm run benchmark:adje
-```
-
 ## Benchmark Result (Latest Local Run)
 
-Benchmark command:
-
-```bash
-npm run benchmark
-```
-
 Results:
 
 - lemma from conjugation: 97/100, accuracy 97.00%, average 21.97 ms

package/package.json

@@ -1,58 +1,52 @@
-{
-  "name": "fr-spell",
-  "description": "Lightweight French lemmatization and conjugation. Convert between nouns, verbs, and adjectives using fast, quantized INT8 ONNX models under 2MB.",
-  "version": "1.0.8",
-  "type": "module",
-  "main": "./dist/index.js",
-  "exports": {
-    ".": "./dist/index.js"
-  },
-  "files": [
-    "dist"
-  ],
-  "scripts": {
-    "help": "node scripts/help.js",
-    "build": "node scripts/build.js",
-    "test": "node test/test.js",
-    "benchmark:prepare": "node benchmark/generate-checklists.js",
-    "benchmark": "node benchmark/run-benchmark.js",
-    "benchmark:lemma": "node benchmark/run-benchmark.js lemma",
-    "benchmark:noun": "node benchmark/run-benchmark.js noun",
-    "benchmark:verb": "node benchmark/run-benchmark.js verb",
-    "benchmark:adje": "node benchmark/run-benchmark.js adje"
-  },
-  "repository": {
-    "type": "git",
-    "url": "https://github.com/davychxn/FR-SPELL"
-  },
-  "keywords": [
-    "french",
-    "Le français",
-    "word lemma",
-    "word derivative",
-    "lemmatization",
-    "conjugation",
-    "spelling",
-    "french noun",
-    "french verb",
-    "french adjective"
-  ],
-  "homepage": "https://coderhome.com/",
-  "license": "MIT",
-  "funding": {
-    "type": "patreon",
-    "url": "https://www.patreon.com/davychxn"
-  },
-  "author": {
-    "name": "Davy Chen",
-    "email": "davy.chen@163.com",
-    "url": "https://www.linkedin.com/in/davychxn/"
-  },
-  "dependencies": {
-    "onnxruntime-node": "^1.24.3",
-    "onnxruntime-web": "^1.24.3"
-  },
-  "devDependencies": {
-    "esbuild": "^0.25.12"
-  }
-}
+{
+  "name": "fr-spell",
+  "version": "1.0.9",
+  "description": "Lightweight French lemmatization and conjugation. Convert between nouns, verbs, and adjectives using fast, quantized INT8 ONNX models under 2MB.",
+  "type": "module",
+  "main": "./index.js",
+  "exports": {
+    ".": "./index.js",
+    "./browser": "./frspell.browser.js"
+  },
+  "files": [
+    "index.js",
+    "frspell.browser.js",
+    "module",
+    "models",
+    "README.md",
+    "README.cn.md",
+    "README.fr.md",
+    "LICENSE"
+  ],
+  "repository": {
+    "type": "git",
+    "url": "https://github.com/davychxn/FR-SPELL"
+  },
+  "keywords": [
+    "french",
+    "Le français",
+    "word lemma",
+    "word derivative",
+    "lemmatization",
+    "conjugation",
+    "spelling",
+    "french noun",
+    "french verb",
+    "french adjective"
+  ],
+  "homepage": "https://coderhome.com/",
+  "license": "MIT",
+  "funding": {
+    "type": "patreon",
+    "url": "https://www.patreon.com/davychxn"
+  },
+  "author": {
+    "name": "Davy Chen",
+    "email": "davy.chen@163.com",
+    "url": "https://www.linkedin.com/in/davychxn/"
+  },
+  "dependencies": {
+    "onnxruntime-node": "^1.24.3",
+    "onnxruntime-web": "^1.24.3"
+  }
+}

package/dist/LICENSE

@@ -1,21 +0,0 @@
-MIT License
-
-Copyright (c) 2026 Davy Chen
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.

package/dist/README.cn.md

@@ -1,142 +0,0 @@
-# FR-SPELL
-
-[English](README.md) | [中文](README.cn.md) | [Français](README.fr.md)
-
-FR-SPELL 是一个用于法语词形还原与派生形态生成的 npm 包。
-支持能力如下：
-
-- 将动词变位形式预测为词元（lemma）
-- 名词词形生成
-- 形容词词形生成
-- 动词词形生成
-
-该包基于 ONNX Runtime 与 INT8 量化模型，兼顾高速度与小体积。
-
-## 安装
-
-```bash
-npm install fr-spell
-```
-
-## 集成到你的项目
-
-```js
-import { FrSpell } from 'fr-spell';
-
-const predictor = await FrSpell();
-
-const lemma = await predictor.lemma('mangeons');
-const noun = await predictor.nounDerive('chat', 'THD_PLF');
-const adje = await predictor.adjeDerive('beau', 'THD_F');
-const verb = await predictor.verbDerive('manger', 'FST_PL', 'INDI', 'PRES');
-
-console.log(lemma);
-console.log(noun);
-console.log(adje);
-console.log(verb);
-```
-
-示例运行输出：
-
-```txt
-{ input: 'mangeons', lemma: 'manger', wordType: 'VERB', confidence: 0.9965604285, timeMs: 3.89 }
-{ lemma: 'chat', wordType: 'NOUN', person: 'THD_PLF', mode: 'ALL', tense: 'ALL', output: 'chattes', confidence: 0.9997230679, timeMs: 5.06 }
-{ lemma: 'beau', wordType: 'ADJE', person: 'THD_F', mode: 'ALL', tense: 'ALL', output: 'belle', confidence: 0.9999751771, timeMs: 3.08 }
-{ lemma: 'manger', wordType: 'VERB', person: 'FST_PL', mode: 'INDI', tense: 'PRES', output: 'mangeons', confidence: 0.9999864523, timeMs: 4.79 }
-```
-
-
-## 浏览器用法
-
-```html
-<script src="./frspell.browser.js"></script>
-<script>
-	(async () => {
-		const predictor = await window.FrSpell({
-			modelBasePath: './models/community'
-		});
-		const result = await predictor.lemma('mangeons');
-		console.log(result);
-	})();
-</script>
-```
-## 预测参数说明
-
-词元预测（lemma）：
-
-- API：`predictor.lemma(input)`
-- `input`：字符串，输入变位/屈折后的词形，例如 `mangeons`
-
-派生预测（derive）：
-
-- 名词 API：`predictor.nounDerive(lemma, person)`
-- 形容词 API：`predictor.adjeDerive(lemma, person)`
-- 动词 API：`predictor.verbDerive(lemma, person, mode, tense)`
-- 通用 API：`predictor.derive(lemma, wordType, person, mode, tense)`
-
-可用 `wordType`：
-
-- `NOUN`（名词）
-- `ADJE`（形容词）
-- `VERB`（动词）
-
-可用 `person`：
-
-- `FST`（第一人称单数）
-- `SND`（第二人称单数）
-- `THD_M`（第三人称阳性单数）
-- `THD_F`（第三人称阴性单数）
-- `FST_PL`（第一人称复数）
-- `SND_PL`（第二人称复数）
-- `THD_PLM`（第三人称阳性复数）
-- `THD_PLF`（第三人称阴性复数）
-
-可用 `mode`：
-
-- `INDI`（陈述式）
-- `SUBJ`（虚拟式）
-- `COND`（条件式）
-- `PART`（分词式）
-- `IMPE`（命令式）
-- `INFI`（不定式）
-
-当前实现支持的 `tense` 仅有：
-
-- `PRES`（现在时）
-- `IMPA`（未完成过去时）
-- `FUTU`（将来时）
-- `PASS`（过去时）
-
-说明：
-
-- 参考定义文件中还有更多时态名称，但本包实现目前只支持 `PRES`、`IMPA`、`FUTU`、`PASS`。
-- 对于名词/形容词派生，用户输入时不需要 `mode` 与 `tense`。
-
-## 基准结果（最近一次本地运行）
-
-结果：
-
-- 由变位预测词元：97/100，准确率 97.00%，平均 21.97 ms
-- 名词派生：100/100，准确率 100.00%，平均 23.19 ms
-- 动词派生：100/100，准确率 100.00%，平均 22.93 ms
-- 形容词派生：100/100，准确率 100.00%，平均 23.22 ms
-
-## 模型体积
-
-- 当前默认（community）词元 ONNX 模型：models/community/lemma_type_model.int8.onnx = 1.48 MB
-- 当前默认（community）派生 ONNX 模型：models/community/derive_form_model.int8.onnx = 1.40 MB
-- 当前默认 ONNX 总体积：约 2.88 MB
-
-Mini 版本说明：
-
-- mini 词元 ONNX 模型目标大小：0.96 MB
-- mini 派生 ONNX 模型目标大小：0.91 MB
-- mini ONNX 总体积目标：约 1.87 MB
-- mini 模型版本将很快上线。
-
-## 为什么它非常适合 Web 前端产品
-
-- 法语关键形态任务具有高准确率
-- 单次请求延迟低（最新本地基准平均约 22 到 23 ms）
-- 当前默认 ONNX 体积紧凑（约 2.88 MB），更小的 mini 模型包（约 1.87 MB）即将上线
-- 非常适合为 Web 前端功能提供后端推理能力，例如实时写作辅助、语法提示与词元感知检索

package/dist/README.fr.md

@@ -1,142 +0,0 @@
-# FR-SPELL
-
-[English](README.md) | [中文](README.cn.md) | [Français](README.fr.md)
-
-FR-SPELL est un package npm pour la prédiction de lemme en français et la génération de formes dérivées.
-Fonctionnalités prises en charge :
-
-- prédiction du lemme à partir d'une forme conjuguée
-- génération de formes nominales
-- génération de formes adjectivales
-- génération de formes verbales
-
-Le package s'appuie sur ONNX Runtime et des modèles INT8 quantifiés pour offrir une grande vitesse avec une empreinte mémoire réduite.
-
-## Installation
-
-```bash
-npm install fr-spell
-```
-
-## Intégration dans votre projet
-
-```js
-import { FrSpell } from 'fr-spell';
-
-const predictor = await FrSpell();
-
-const lemma = await predictor.lemma('mangeons');
-const noun = await predictor.nounDerive('chat', 'THD_PLF');
-const adje = await predictor.adjeDerive('beau', 'THD_F');
-const verb = await predictor.verbDerive('manger', 'FST_PL', 'INDI', 'PRES');
-
-console.log(lemma);
-console.log(noun);
-console.log(adje);
-console.log(verb);
-```
-
-Exemple de sortie a l'execution :
-
-```txt
-{ input: 'mangeons', lemma: 'manger', wordType: 'VERB', confidence: 0.9965604285, timeMs: 3.89 }
-{ lemma: 'chat', wordType: 'NOUN', person: 'THD_PLF', mode: 'ALL', tense: 'ALL', output: 'chattes', confidence: 0.9997230679, timeMs: 5.06 }
-{ lemma: 'beau', wordType: 'ADJE', person: 'THD_F', mode: 'ALL', tense: 'ALL', output: 'belle', confidence: 0.9999751771, timeMs: 3.08 }
-{ lemma: 'manger', wordType: 'VERB', person: 'FST_PL', mode: 'INDI', tense: 'PRES', output: 'mangeons', confidence: 0.9999864523, timeMs: 4.79 }
-```
-
-
-## Utilisation dans le navigateur
-
-```html
-<script src="./frspell.browser.js"></script>
-<script>
-	(async () => {
-		const predictor = await window.FrSpell({
-			modelBasePath: './models/community'
-		});
-		const result = await predictor.lemma('mangeons');
-		console.log(result);
-	})();
-</script>
-```
-## Parametres de prediction
-
-Prediction de lemme :
-
-- API : `predictor.lemma(input)`
-- `input` : chaine de caracteres, forme flechie/conjuguee, par exemple `mangeons`
-
-Prediction de derive :
-
-- API nom : `predictor.nounDerive(lemma, person)`
-- API adjectif : `predictor.adjeDerive(lemma, person)`
-- API verbe : `predictor.verbDerive(lemma, person, mode, tense)`
-- API generique : `predictor.derive(lemma, wordType, person, mode, tense)`
-
-Valeurs `wordType` autorisees :
-
-- `NOUN` (nom)
-- `ADJE` (adjectif)
-- `VERB` (verbe)
-
-Valeurs `person` autorisees :
-
-- `FST` (1re personne du singulier)
-- `SND` (2e personne du singulier)
-- `THD_M` (3e personne masculine du singulier)
-- `THD_F` (3e personne feminine du singulier)
-- `FST_PL` (1re personne du pluriel)
-- `SND_PL` (2e personne du pluriel)
-- `THD_PLM` (3e personne masculine du pluriel)
-- `THD_PLF` (3e personne feminine du pluriel)
-
-Valeurs `mode` autorisees :
-
-- `INDI` (indicatif)
-- `SUBJ` (subjonctif)
-- `COND` (conditionnel)
-- `PART` (participe)
-- `IMPE` (imperatif)
-- `INFI` (infinitif)
-
-Valeurs `tense` prises en charge dans l'implementation actuelle :
-
-- `PRES` (present)
-- `IMPA` (imparfait)
-- `FUTU` (futur)
-- `PASS` (passe)
-
-Note :
-
-- Le fichier de definitions d'origine contient plus de noms de temps, mais ce package prend actuellement en charge uniquement `PRES`, `IMPA`, `FUTU`, `PASS`.
-- Pour les appels nom/adjectif, `mode` et `tense` ne sont pas requis dans l'entree utilisateur.
-
-## Résultats de benchmark (dernier run local)
-
-Résultats :
-
-- lemme depuis une conjugaison : 97/100, précision 97.00 %, moyenne 21.97 ms
-- dérivation nominale : 100/100, précision 100.00 %, moyenne 23.19 ms
-- dérivation verbale : 100/100, précision 100.00 %, moyenne 22.93 ms
-- dérivation adjectivale : 100/100, précision 100.00 %, moyenne 23.22 ms
-
-## Taille des modèles
-
-- modèle ONNX lemme par défaut (community) : models/community/lemma_type_model.int8.onnx = 1.48 MB
-- modèle ONNX dérivation par défaut (community) : models/community/derive_form_model.int8.onnx = 1.40 MB
-- taille ONNX totale par défaut actuelle : environ 2.88 MB
-
-Note sur la version mini :
-
-- taille cible du modèle lemme mini : 0.96 MB
-- taille cible du modèle dérivation mini : 0.91 MB
-- taille ONNX totale cible de la version mini : environ 1.87 MB
-- le package de modèles mini sera publié prochainement.
-
-## Pourquoi c'est idéal pour des produits web frontend
-
-- excellente précision sur les tâches clés de morphologie française
-- faible latence par requête (environ 22 à 23 ms en moyenne sur le dernier benchmark local)
-- empreinte ONNX par défaut toujours compacte (environ 2.88 MB au total), avec un package mini plus léger (environ 1.87 MB) à venir
-- parfait pour alimenter des fonctionnalités frontend via une inférence backend : assistance à l'écriture en temps réel, suggestions grammaticales et recherche basée sur le lemme

package/dist/README.md

@@ -1,142 +0,0 @@
-# FR-SPELL
-
-[English](README.md) | [中文](README.cn.md) | [Français](README.fr.md)
-
-FR-SPELL is an npm package for French lemma prediction and derivative form generation.
-It supports:
-
-- conjugation to lemma prediction
-- noun form generation
-- adjective form generation
-- verb form generation
-
-The package runs with ONNX Runtime and quantized INT8 models for high speed and small model footprint.
-
-## Install
-
-```bash
-npm install fr-spell
-```
-
-## Integrate Into Your Project
-
-```js
-import { FrSpell } from 'fr-spell';
-
-const predictor = await FrSpell();
-
-const lemma = await predictor.lemma('mangeons');
-const noun = await predictor.nounDerive('chat', 'THD_PLF');
-const adje = await predictor.adjeDerive('beau', 'THD_F');
-const verb = await predictor.verbDerive('manger', 'FST_PL', 'INDI', 'PRES');
-
-console.log(lemma);
-console.log(noun);
-console.log(adje);
-console.log(verb);
-```
-
-Sample runtime output:
-
-```txt
-{ input: 'mangeons', lemma: 'manger', wordType: 'VERB', confidence: 0.9965604285, timeMs: 3.89 }
-{ lemma: 'chat', wordType: 'NOUN', person: 'THD_PLF', mode: 'ALL', tense: 'ALL', output: 'chattes', confidence: 0.9997230679, timeMs: 5.06 }
-{ lemma: 'beau', wordType: 'ADJE', person: 'THD_F', mode: 'ALL', tense: 'ALL', output: 'belle', confidence: 0.9999751771, timeMs: 3.08 }
-{ lemma: 'manger', wordType: 'VERB', person: 'FST_PL', mode: 'INDI', tense: 'PRES', output: 'mangeons', confidence: 0.9999864523, timeMs: 4.79 }
-```
-
-
-## Browser Usage
-
-```html
-<script src="./frspell.browser.js"></script>
-<script>
-	(async () => {
-		const predictor = await window.FrSpell({
-			modelBasePath: './models/community'
-		});
-		const result = await predictor.lemma('mangeons');
-		console.log(result);
-	})();
-</script>
-```
-## Prediction Parameters
-
-Lemma prediction:
-
-- API: `predictor.lemma(input)`
-- `input`: string, inflected/conjugated word form, for example `mangeons`
-
-Derive prediction:
-
-- Noun API: `predictor.nounDerive(lemma, person)`
-- Adjective API: `predictor.adjeDerive(lemma, person)`
-- Verb API: `predictor.verbDerive(lemma, person, mode, tense)`
-- Generic API: `predictor.derive(lemma, wordType, person, mode, tense)`
-
-Allowed `wordType` values:
-
-- `NOUN` (noun)
-- `ADJE` (adjective)
-- `VERB` (verb)
-
-Allowed `person` values:
-
-- `FST` (1st person singular)
-- `SND` (2nd person singular)
-- `THD_M` (3rd person masculine singular)
-- `THD_F` (3rd person feminine singular)
-- `FST_PL` (1st person plural)
-- `SND_PL` (2nd person plural)
-- `THD_PLM` (3rd person masculine plural)
-- `THD_PLF` (3rd person feminine plural)
-
-Allowed `mode` values:
-
-- `INDI` (indicative)
-- `SUBJ` (subjunctive)
-- `COND` (conditional)
-- `PART` (participle)
-- `IMPE` (imperative)
-- `INFI` (infinitive)
-
-Allowed `tense` values in current implementation:
-
-- `PRES` (present)
-- `IMPA` (imperfect)
-- `FUTU` (future)
-- `PASS` (past)
-
-Note:
-
-- The original grammar definition file includes more tense names, but this package implementation currently supports only `PRES`, `IMPA`, `FUTU`, `PASS`.
-- For noun/adjective derive calls, `mode` and `tense` are not required in user input.
-
-## Benchmark Result (Latest Local Run)
-
-Results:
-
-- lemma from conjugation: 97/100, accuracy 97.00%, average 21.97 ms
-- noun derive: 100/100, accuracy 100.00%, average 23.19 ms
-- verb derive: 100/100, accuracy 100.00%, average 22.93 ms
-- adjective derive: 100/100, accuracy 100.00%, average 23.22 ms
-
-## Model Size
-
-- current default (community) lemma ONNX model: models/community/lemma_type_model.int8.onnx = 1.48 MB
-- current default (community) derive ONNX model: models/community/derive_form_model.int8.onnx = 1.40 MB
-- current default total ONNX model size: about 2.88 MB
-
-Mini version note:
-
-- mini lemma ONNX model target: 0.96 MB
-- mini derive ONNX model target: 0.91 MB
-- mini total ONNX model target: about 1.87 MB
-- the mini model package is planned to be published soon.
-
-## Why It Is Great For Web Frontend Products
-
-- high accuracy for key French morphology tasks
-- low per-request latency (about 22 to 23 ms average in latest local benchmark)
-- current default ONNX footprint is compact (about 2.88 MB total), with a smaller mini model package (about 1.87 MB) coming soon
-- ideal for backend inference powering web frontend features such as live writing assistance, grammar hints, and lemma-aware search

package/dist/package.json

@@ -1,52 +0,0 @@
-{
-  "name": "fr-spell",
-  "version": "1.0.6",
-  "description": "Lightweight French lemmatization and conjugation. Convert between nouns, verbs, and adjectives using fast, quantized INT8 ONNX models under 2MB.",
-  "type": "module",
-  "main": "./index.js",
-  "exports": {
-    ".": "./index.js",
-    "./browser": "./frspell.browser.js"
-  },
-  "files": [
-    "index.js",
-    "frspell.browser.js",
-    "module",
-    "models",
-    "README.md",
-    "README.cn.md",
-    "README.fr.md",
-    "LICENSE"
-  ],
-  "repository": {
-    "type": "git",
-    "url": "https://github.com/davychxn/FR-SPELL"
-  },
-  "keywords": [
-    "french",
-    "Le français",
-    "word lemma",
-    "word derivative",
-    "lemmatization",
-    "conjugation",
-    "spelling",
-    "french noun",
-    "french verb",
-    "french adjective"
-  ],
-  "homepage": "https://github.com/davychxn/FR-SPELL",
-  "license": "MIT",
-  "funding": {
-    "type": "patreon",
-    "url": "https://www.patreon.com/davychxn"
-  },
-  "author": {
-    "name": "Davy Chen",
-    "email": "davy.chen@163.com",
-    "url": "https://www.linkedin.com/in/davychxn/"
-  },
-  "dependencies": {
-    "onnxruntime-node": "^1.24.3",
-    "onnxruntime-web": "^1.24.3"
-  }
-}