@orkg/scidquest 1.0.0

package/dist/semanticChunker-CEckDeUn.js

@@ -0,0 +1,102 @@
+var g = Object.defineProperty;
+var p = (a, e, n) => e in a ? g(a, e, { enumerable: !0, configurable: !0, writable: !0, value: n }) : a[e] = n;
+var u = (a, e, n) => p(a, typeof e != "symbol" ? e + "" : e, n);
+import { pipeline as k } from "@huggingface/transformers";
+class f {
+  constructor() {
+    u(this, "model", null);
+    u(this, "modelLoading", null);
+  }
+  async initialize() {
+    if (!this.model) {
+      if (this.modelLoading) return this.modelLoading;
+      console.log("[FrontendSemanticChunker] Initializing model..."), this.modelLoading = (async () => {
+        try {
+          this.model = await k(
+            "feature-extraction",
+            "Xenova/all-MiniLM-L6-v2",
+            { dtype: "q8" }
+          ), console.log("[FrontendSemanticChunker] Model loaded successfully");
+        } catch (e) {
+          throw console.error("[FrontendSemanticChunker] Model initialization failed:", e), this.modelLoading = null, e;
+        }
+      })(), await this.modelLoading;
+    }
+  }
+  splitIntoChunks(e) {
+    const n = [];
+    for (const s of e) {
+      if (!s.text || s.text.trim().length === 0) continue;
+      const o = s.text.split(new RegExp("(?<=[.!?])(?=\\s)"));
+      let t = "", r = 0;
+      for (const h of o) {
+        const m = h.trim();
+        m && (t && t.length + m.length + 1 > 2e3 ? (n.push({
+          text: t.trim(),
+          pageNumber: s.pageNumber,
+          startIndex: r,
+          endIndex: r + t.length,
+          tokenEstimate: Math.ceil(t.length / 4)
+        }), t = t.trim().split(/\s+/).slice(-Math.ceil(400 / 5)).join(" ") + " " + m, r += t.length - 400) : t += (t ? " " : "") + m);
+      }
+      t.trim() && (n.push({
+        text: t.trim(),
+        pageNumber: s.pageNumber,
+        startIndex: r,
+        endIndex: r + t.length,
+        tokenEstimate: Math.ceil(t.length / 4)
+      }), t = "", r = 0);
+    }
+    return console.log(
+      `[FrontendSemanticChunker] Total chunks: ${n.length}, Target size: 2000 chars`
+    ), n;
+  }
+  async embed(e) {
+    const n = await this.model(e, { pooling: "mean", normalize: !0 });
+    return Array.from(n.data);
+  }
+  cosineSimilarity(e, n) {
+    let c = 0, l = 0, s = 0;
+    for (let o = 0; o < e.length; o++)
+      c += e[o] * n[o], l += e[o] * e[o], s += n[o] * n[o];
+    return c / (Math.sqrt(l) * Math.sqrt(s));
+  }
+  async findRelevantChunks(e, n, c = 4e3) {
+    console.log(
+      `[FrontendSemanticChunker] Finding relevant chunks for question (max ${c} tokens)`
+    ), await this.initialize();
+    const l = this.splitIntoChunks(n);
+    console.log(
+      `[FrontendSemanticChunker] Created ${l.length} chunks from ${n.length} pages`
+    ), console.log("[FrontendSemanticChunker] Generating embeddings...");
+    const s = await Promise.all(
+      l.map((i) => this.embed(i.text))
+    ), o = await this.embed(e), t = l.map((i, d) => ({
+      ...i,
+      similarity: this.cosineSimilarity(o, s[d])
+    }));
+    t.sort((i, d) => d.similarity - i.similarity), console.log(
+      "[FrontendSemanticChunker] Top 5 similarity scores:",
+      t.slice(0, 5).map((i) => `page ${i.pageNumber}: ${i.similarity.toFixed(4)}`)
+    );
+    const r = [];
+    let h = 0;
+    for (const i of t)
+      h + i.tokenEstimate <= c && (r.push(i), h += i.tokenEstimate);
+    console.log(
+      `[FrontendSemanticChunker] Selected ${r.length} chunks (${h} tokens)`
+    ), r.sort((i, d) => i.pageNumber - d.pageNumber);
+    const m = Array.from(new Set(r.map((i) => i.pageNumber)));
+    return console.log(
+      "[FrontendSemanticChunker] Chunks span pages:",
+      m.join(", ")
+    ), r;
+  }
+  isInitialized() {
+    return this.model !== null;
+  }
+}
+const y = new f();
+export {
+  y as frontendSemanticChunker
+};

package/package.json

@@ -0,0 +1,63 @@
+{
+  "name": "@orkg/scidquest",
+  "version": "1.0.0",
+  "type": "module",
+  "description": "Standalone research paper contribution and analysis tool",
+  "main": "./dist/scidquest.es.js",
+  "module": "./dist/scidquest.es.js",
+  "types": "./dist/index.d.ts",
+  "files": [
+    "dist"
+  ],
+  "exports": {
+    ".": {
+      "types": "./dist/index.d.ts",
+      "import": "./dist/scidquest.es.js",
+      "default": "./dist/scidquest.es.js"
+    },
+    "./dist/contribute-standalone.css": "./dist/contribute-standalone.css"
+  },
+  "scripts": {
+    "dev": "vite",
+    "build": "tsc && vite build",
+    "preview": "vite preview",
+    "prepublishOnly": "npm run build"
+  },
+  "peerDependencies": {
+    "@emotion/react": "^11.14.0",
+    "@emotion/styled": "^11.14.0",
+    "@mui/material": "^6.4.11",
+    "react": "^18.3.1",
+    "react-dom": "^18.3.1"
+  },
+  "dependencies": {
+    "@ai-sdk/groq": "^1.2.9",
+    "@ai-sdk/mistral": "^2.0.23",
+    "@ai-sdk/openai": "^2.0.0-alpha.3",
+    "@emotion/react": "^11.14.0",
+    "@emotion/styled": "^11.14.0",
+    "@huggingface/transformers": "^4.2.0",
+    "@mui/icons-material": "^6.4.10",
+    "@mui/material": "^6.4.11",
+    "@reduxjs/toolkit": "^2.8.2",
+    "ai": "^5.0.86",
+    "pdfjs-dist": "^5.4.296",
+    "react": "^18.3.1",
+    "react-dom": "^18.3.1",
+    "react-pdf": "^10.3.0",
+    "react-redux": "^9.2.0",
+    "react-router-dom": "^7.5.0"
+  },
+  "devDependencies": {
+    "@types/node": "^25.1.0",
+    "@types/react": "^18.3.18",
+    "@types/react-dom": "^18.3.5",
+    "@vitejs/plugin-react": "^4.3.4",
+    "typescript": "~5.6.2",
+    "vite": "^6.0.5",
+    "vite-plugin-static-copy": "^4.1.0"
+  },
+  "publishConfig": {
+    "access": "public"
+  }
+}