@yuaone/core 0.9.13 → 0.9.15

package/dist/vision-intent-detector.js

@@ -0,0 +1,241 @@
+/**
+ * @module vision-intent-detector
+ * @description Detects when LLM or user text signals intent to look at an image,
+ * then extracts the target file path. Supports 9+ languages.
+ */
+const LANG_PATTERNS = [
+    {
+        lang: "ko",
+        phrases: [
+            "이미지 확인",
+            "스크린샷 확인",
+            "사진 봐",
+            "그림 확인",
+            "이미지 봐",
+            "화면 봐",
+            "파일 읽어",
+            "이미지 읽어",
+            "스크린샷 읽어",
+            "이미지를 확인",
+            "사진을 봐",
+            "그림을 확인",
+            "이미지를 봐",
+            "화면을 봐",
+        ],
+    },
+    {
+        lang: "en",
+        phrases: [
+            "let me see",
+            "let me look at",
+            "check the image",
+            "look at",
+            "view the screenshot",
+            "analyze the image",
+            "read the image",
+            "show me",
+            "what's in",
+            "examine",
+            "inspect the image",
+            "open the image",
+            "view the image",
+            "check the screenshot",
+            "analyze the screenshot",
+        ],
+    },
+    {
+        lang: "ja",
+        phrases: [
+            "画像を確認",
+            "スクリーンショットを見",
+            "イメージを確認",
+            "画像を見て",
+            "画像を読む",
+            "画像を調べる",
+        ],
+    },
+    {
+        lang: "zh",
+        phrases: [
+            "查看图片",
+            "看一下图",
+            "检查截图",
+            "分析图像",
+            "查看截图",
+            "看看图片",
+            "读取图片",
+        ],
+    },
+    {
+        lang: "es",
+        phrases: [
+            "ver la imagen",
+            "analizar imagen",
+            "revisar captura",
+            "mirar la imagen",
+            "examinar la imagen",
+            "ver captura de pantalla",
+        ],
+    },
+    {
+        lang: "fr",
+        phrases: [
+            "voir l'image",
+            "regarder l'image",
+            "analyser l'image",
+            "vérifier l'image",
+            "examiner l'image",
+            "regarder la capture",
+        ],
+    },
+    {
+        lang: "de",
+        phrases: [
+            "bild ansehen",
+            "screenshot prüfen",
+            "bild analysieren",
+            "bild überprüfen",
+            "screenshot ansehen",
+            "bild lesen",
+        ],
+    },
+    {
+        lang: "ru",
+        phrases: [
+            "посмотрите на изображение",
+            "проверь скриншот",
+            "посмотри на изображение",
+            "проверить скриншот",
+            "посмотри на картинку",
+            "проверь изображение",
+        ],
+    },
+    {
+        lang: "ar",
+        phrases: [
+            "انظر إلى الصورة",
+            "تحقق من الصورة",
+            "راجع الصورة",
+            "حلل الصورة",
+            "فحص لقطة الشاشة",
+        ],
+    },
+];
+// ─── File path extraction regex ─────────────────────────────────────────────
+/** Matches quoted image paths: "foo.png", 'bar.jpg', `baz.webp` */
+const QUOTED_IMAGE_PATH_RE = /['"`]([^'"`\s]+\.(?:png|jpg|jpeg|gif|webp|svg|bmp))['"`]/gi;
+/** Matches bare image paths (relative or absolute, no spaces) */
+const BARE_IMAGE_PATH_RE = /(?:^|[\s(,])(\/?(?:[\w.\-]+\/)*[\w.\-]+\.(?:png|jpg|jpeg|gif|webp|svg|bmp))(?:$|[\s),])/gi;
+// ─── Language detection helpers ──────────────────────────────────────────────
+function detectScript(text) {
+    if (/[\uAC00-\uD7A3]/.test(text))
+        return "ko";
+    if (/[\u3040-\u309F\u30A0-\u30FF]/.test(text))
+        return "ja";
+    if (/[\u4E00-\u9FFF]/.test(text))
+        return "zh";
+    if (/[\u0600-\u06FF]/.test(text))
+        return "ar";
+    if (/[\u0400-\u04FF]/.test(text))
+        return "ru";
+    return null;
+}
+// ─── Core class ─────────────────────────────────────────────────────────────
+/**
+ * Detects vision intent in LLM / user text and extracts image file paths.
+ *
+ * @example
+ * ```ts
+ * const detector = new VisionIntentDetector();
+ * const intent = detector.detect('let me look at "screenshot.png" to diagnose the error');
+ * // → { detectedLanguage: "en", filePath: "screenshot.png", intentPhrase: "let me look at", confidence: 0.95 }
+ * ```
+ */
+export class VisionIntentDetector {
+    /**
+     * Attempt to detect a vision intent in `text`.
+     *
+     * Returns `null` when no intent signal is found, or when an intent is found
+     * but no image file path can be extracted from the surrounding context.
+     */
+    detect(text) {
+        if (!text || text.trim().length === 0)
+            return null;
+        const lower = text.toLowerCase();
+        // Try each language's patterns
+        for (const { lang, phrases } of LANG_PATTERNS) {
+            for (const phrase of phrases) {
+                const phraseToCheck = lang === "de" ? phrase.toLowerCase() : phrase;
+                if (lower.includes(phraseToCheck)) {
+                    // Phrase matched — now find an image path in the text
+                    const filePath = extractImagePath(text);
+                    if (filePath) {
+                        // Higher confidence when the phrase is more specific
+                        const confidence = computeConfidence(phrase, filePath);
+                        return {
+                            detectedLanguage: lang,
+                            filePath,
+                            intentPhrase: phrase,
+                            confidence,
+                        };
+                    }
+                }
+            }
+        }
+        // Fallback: detect by Unicode script only (phrase not in list but script is clear)
+        const scriptLang = detectScript(text);
+        if (scriptLang) {
+            const filePath = extractImagePath(text);
+            if (filePath) {
+                return {
+                    detectedLanguage: scriptLang,
+                    filePath,
+                    intentPhrase: "",
+                    confidence: 0.4, // low confidence — no phrase match
+                };
+            }
+        }
+        return null;
+    }
+}
+// ─── Helpers ─────────────────────────────────────────────────────────────────
+/**
+ * Extract the first image file path from text.
+ * Prefers quoted paths (more explicit) over bare paths.
+ */
+function extractImagePath(text) {
+    // Reset lastIndex (global regex reuse safety)
+    QUOTED_IMAGE_PATH_RE.lastIndex = 0;
+    BARE_IMAGE_PATH_RE.lastIndex = 0;
+    const quotedMatch = QUOTED_IMAGE_PATH_RE.exec(text);
+    if (quotedMatch?.[1])
+        return quotedMatch[1];
+    const bareMatch = BARE_IMAGE_PATH_RE.exec(text);
+    if (bareMatch?.[1])
+        return bareMatch[1].trim();
+    return null;
+}
+/**
+ * Compute confidence score based on phrase specificity and path quality.
+ *
+ * - Long, specific phrases (e.g. "analyze the image") → 0.90–0.95
+ * - Short generic phrases (e.g. "show me") → 0.70–0.80
+ * - Quoted path found → +0.05 bonus (capped at 1.0)
+ */
+function computeConfidence(phrase, filePath) {
+    let base = 0.75;
+    // Longer phrase → more specific intent
+    if (phrase.length >= 15)
+        base = 0.95;
+    else if (phrase.length >= 10)
+        base = 0.90;
+    else if (phrase.length >= 6)
+        base = 0.80;
+    // Quoted path → more explicit reference
+    if (/['"`]/.test(filePath)) {
+        // path itself was extracted from quotes
+        base = Math.min(1.0, base + 0.05);
+    }
+    return Math.round(base * 100) / 100;
+}
+//# sourceMappingURL=vision-intent-detector.js.map

package/dist/vision-intent-detector.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"vision-intent-detector.js","sourceRoot":"","sources":["../src/vision-intent-detector.ts"],"names":[],"mappings":"AAAA;;;;GAIG;AAoBH,MAAM,aAAa,GAAkB;IACnC;QACE,IAAI,EAAE,IAAI;QACV,OAAO,EAAE;YACP,QAAQ;YACR,SAAS;YACT,MAAM;YACN,OAAO;YACP,OAAO;YACP,MAAM;YACN,OAAO;YACP,QAAQ;YACR,SAAS;YACT,SAAS;YACT,OAAO;YACP,QAAQ;YACR,QAAQ;YACR,OAAO;SACR;KACF;IACD;QACE,IAAI,EAAE,IAAI;QACV,OAAO,EAAE;YACP,YAAY;YACZ,gBAAgB;YAChB,iBAAiB;YACjB,SAAS;YACT,qBAAqB;YACrB,mBAAmB;YACnB,gBAAgB;YAChB,SAAS;YACT,WAAW;YACX,SAAS;YACT,mBAAmB;YACnB,gBAAgB;YAChB,gBAAgB;YAChB,sBAAsB;YACtB,wBAAwB;SACzB;KACF;IACD;QACE,IAAI,EAAE,IAAI;QACV,OAAO,EAAE;YACP,OAAO;YACP,aAAa;YACb,SAAS;YACT,OAAO;YACP,OAAO;YACP,QAAQ;SACT;KACF;IACD;QACE,IAAI,EAAE,IAAI;QACV,OAAO,EAAE;YACP,MAAM;YACN,MAAM;YACN,MAAM;YACN,MAAM;YACN,MAAM;YACN,MAAM;YACN,MAAM;SACP;KACF;IACD;QACE,IAAI,EAAE,IAAI;QACV,OAAO,EAAE;YACP,eAAe;YACf,iBAAiB;YACjB,iBAAiB;YACjB,iBAAiB;YACjB,oBAAoB;YACpB,yBAAyB;SAC1B;KACF;IACD;QACE,IAAI,EAAE,IAAI;QACV,OAAO,EAAE;YACP,cAAc;YACd,kBAAkB;YAClB,kBAAkB;YAClB,kBAAkB;YAClB,kBAAkB;YAClB,qBAAqB;SACtB;KACF;IACD;QACE,IAAI,EAAE,IAAI;QACV,OAAO,EAAE;YACP,cAAc;YACd,mBAAmB;YACnB,kBAAkB;YAClB,iBAAiB;YACjB,oBAAoB;YACpB,YAAY;SACb;KACF;IACD;QACE,IAAI,EAAE,IAAI;QACV,OAAO,EAAE;YACP,2BAA2B;YAC3B,kBAAkB;YAClB,yBAAyB;YACzB,oBAAoB;YACpB,sBAAsB;YACtB,qBAAqB;SACtB;KACF;IACD;QACE,IAAI,EAAE,IAAI;QACV,OAAO,EAAE;YACP,iBAAiB;YACjB,gBAAgB;YAChB,aAAa;YACb,YAAY;YACZ,iBAAiB;SAClB;KACF;CACF,CAAC;AAEF,+EAA+E;AAE/E,mEAAmE;AACnE,MAAM,oBAAoB,GACxB,4DAA4D,CAAC;AAE/D,iEAAiE;AACjE,MAAM,kBAAkB,GACtB,2FAA2F,CAAC;AAE9F,gFAAgF;AAEhF,SAAS,YAAY,CAAC,IAAY;IAChC,IAAI,iBAAiB,CAAC,IAAI,CAAC,IAAI,CAAC;QAAE,OAAO,IAAI,CAAC;IAC9C,IAAI,8BAA8B,CAAC,IAAI,CAAC,IAAI,CAAC;QAAE,OAAO,IAAI,CAAC;IAC3D,IAAI,iBAAiB,CAAC,IAAI,CAAC,IAAI,CAAC;QAAE,OAAO,IAAI,CAAC;IAC9C,IAAI,iBAAiB,CAAC,IAAI,CAAC,IAAI,CAAC;QAAE,OAAO,IAAI,CAAC;IAC9C,IAAI,iBAAiB,CAAC,IAAI,CAAC,IAAI,CAAC;QAAE,OAAO,IAAI,CAAC;IAC9C,OAAO,IAAI,CAAC;AACd,CAAC;AAED,+EAA+E;AAE/E;;;;;;;;;GASG;AACH,MAAM,OAAO,oBAAoB;IAC/B;;;;;OAKG;IACH,MAAM,CAAC,IAAY;QACjB,IAAI,CAAC,IAAI,IAAI,IAAI,CAAC,IAAI,EAAE,CAAC,MAAM,KAAK,CAAC;YAAE,OAAO,IAAI,CAAC;QAEnD,MAAM,KAAK,GAAG,IAAI,CAAC,WAAW,EAAE,CAAC;QAEjC,+BAA+B;QAC/B,KAAK,MAAM,EAAE,IAAI,EAAE,OAAO,EAAE,IAAI,aAAa,EAAE,CAAC;YAC9C,KAAK,MAAM,MAAM,IAAI,OAAO,EAAE,CAAC;gBAC7B,MAAM,aAAa,GAAG,IAAI,KAAK,IAAI,CAAC,CAAC,CAAC,MAAM,CAAC,WAAW,EAAE,CAAC,CAAC,CAAC,MAAM,CAAC;gBACpE,IAAI,KAAK,CAAC,QAAQ,CAAC,aAAa,CAAC,EAAE,CAAC;oBAClC,sDAAsD;oBACtD,MAAM,QAAQ,GAAG,gBAAgB,CAAC,IAAI,CAAC,CAAC;oBACxC,IAAI,QAAQ,EAAE,CAAC;wBACb,qDAAqD;wBACrD,MAAM,UAAU,GAAG,iBAAiB,CAAC,MAAM,EAAE,QAAQ,CAAC,CAAC;wBACvD,OAAO;4BACL,gBAAgB,EAAE,IAAI;4BACtB,QAAQ;4BACR,YAAY,EAAE,MAAM;4BACpB,UAAU;yBACX,CAAC;oBACJ,CAAC;gBACH,CAAC;YACH,CAAC;QACH,CAAC;QAED,mFAAmF;QACnF,MAAM,UAAU,GAAG,YAAY,CAAC,IAAI,CAAC,CAAC;QACtC,IAAI,UAAU,EAAE,CAAC;YACf,MAAM,QAAQ,GAAG,gBAAgB,CAAC,IAAI,CAAC,CAAC;YACxC,IAAI,QAAQ,EAAE,CAAC;gBACb,OAAO;oBACL,gBAAgB,EAAE,UAAU;oBAC5B,QAAQ;oBACR,YAAY,EAAE,EAAE;oBAChB,UAAU,EAAE,GAAG,EAAE,mCAAmC;iBACrD,CAAC;YACJ,CAAC;QACH,CAAC;QAED,OAAO,IAAI,CAAC;IACd,CAAC;CACF;AAED,gFAAgF;AAEhF;;;GAGG;AACH,SAAS,gBAAgB,CAAC,IAAY;IACpC,8CAA8C;IAC9C,oBAAoB,CAAC,SAAS,GAAG,CAAC,CAAC;IACnC,kBAAkB,CAAC,SAAS,GAAG,CAAC,CAAC;IAEjC,MAAM,WAAW,GAAG,oBAAoB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IACpD,IAAI,WAAW,EAAE,CAAC,CAAC,CAAC;QAAE,OAAO,WAAW,CAAC,CAAC,CAAC,CAAC;IAE5C,MAAM,SAAS,GAAG,kBAAkB,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC;IAChD,IAAI,SAAS,EAAE,CAAC,CAAC,CAAC;QAAE,OAAO,SAAS,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;IAE/C,OAAO,IAAI,CAAC;AACd,CAAC;AAED;;;;;;GAMG;AACH,SAAS,iBAAiB,CAAC,MAAc,EAAE,QAAgB;IACzD,IAAI,IAAI,GAAG,IAAI,CAAC;IAEhB,uCAAuC;IACvC,IAAI,MAAM,CAAC,MAAM,IAAI,EAAE;QAAE,IAAI,GAAG,IAAI,CAAC;SAChC,IAAI,MAAM,CAAC,MAAM,IAAI,EAAE;QAAE,IAAI,GAAG,IAAI,CAAC;SACrC,IAAI,MAAM,CAAC,MAAM,IAAI,CAAC;QAAE,IAAI,GAAG,IAAI,CAAC;IAEzC,wCAAwC;IACxC,IAAI,OAAO,CAAC,IAAI,CAAC,QAAQ,CAAC,EAAE,CAAC;QAC3B,wCAAwC;QACxC,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,GAAG,EAAE,IAAI,GAAG,IAAI,CAAC,CAAC;IACpC,CAAC;IAED,OAAO,IAAI,CAAC,KAAK,CAAC,IAAI,GAAG,GAAG,CAAC,GAAG,GAAG,CAAC;AACtC,CAAC"}

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@yuaone/core",
-  "version": "0.9.13",
+  "version": "0.9.15",
   "description": "YUAN Agent Core Runtime",
   "license": "AGPL-3.0",
   "type": "module",