Blame for app.js

import {
  AutoModelForSequenceClassification,
  AutoTokenizer,
} from "https://cdn.jsdelivr.net/npm/@huggingface/transformers@3.8.1/+esm";
import * as pdfjsLib from "https://cdn.jsdelivr.net/npm/pdfjs-dist@4.10.38/build/pdf.min.mjs";

const MODEL_ID = "Gehan77/cv-match-browser";
const DECISION_THRESHOLD = 0.4399277865886688;
pdfjsLib.GlobalWorkerOptions.workerSrc = "https://cdn.jsdelivr.net/npm/pdfjs-dist@4.10.38/build/pdf.worker.min.mjs";

const $ = (id) => document.getElementById(id);
const cvFile = $("cvFile");
const dropzone = $("dropzone");
const fileLabel = $("fileLabel");
const extractButton = $("extractButton");
const scoreButton = $("scoreButton");
const status = $("status");
const preview = $("preview");
const result = $("result");
let selectedFile = null;
let modelPromise = null;

function setStatus(message, type = "") {
  status.textContent = message;
  status.className = `status ${type}`.trim();
}

function clean(value, fallback = "Not specified") {
  const text = String(value ?? "").replace(/\s+/g, " ").trim();
  return text || fallback;
}

function cleanList(value) {
  const text = clean(value);
  if (text === "Not specified") return text;
  const items = text.split(/[|,;\n]+/).map((item) => item.trim()).filter(Boolean);
  return items.length ? items.join(", ") : "Not specified";
}

const headerRules = [
  ["summary", /^(?:(?:professional\s+)?summary|profile|about\s+me|objective|career\s+objective)$/i],
  ["experience", /^(?:(?:professional\s+)?(?:work\s+)?experience|employment\s+history|career\s+history)$/i],
  ["education", /^(?:education(?:al\s+background)?|academic\s+qualifications?)$/i],
  ["skills", /^(?:(?:technical\s+|core\s+|key\s+|professional\s+)?skills?|competenc(?:y|ies)|technologies)$/i],
];

function headerName(line) {
  const candidate = line.replace(/[#:|\-–—]+$/g, "").trim();
  if (candidate.length > 45) return null;
  return headerRules.find(([, pattern]) => pattern.test(candidate))?.[0] ?? null;
}

function parseSections(text) {
  const sections = { other: [] };
  let current = "other";
  for (const rawLine of text.split(/\r?\n/)) {
    const line = rawLine.replace(/^[\s•*\-–—]+/, "").trim();
    if (!line) continue;
    const header = headerName(line);
    if (header) {
      current = header;
      sections[current] ??= [];
      continue;
    }
    sections[current] ??= [];
    sections[current].push(line);
  }
  return sections;
}

function firstNonContactLine(text) {
  for (const rawLine of text.split(/\r?\n/).slice(0, 12)) {
    const line = rawLine.replace(/^[\s•*\-–—]+/, "").trim();
    if (!line || headerName(line)) continue;
    if (/@|https?:\/\/|www\.|(?:\+?\d[\d ()-]{7,})/i.test(line)) continue;
    if (line.split(/\s+/).length <= 10) return line;
  }
  return "";
}

function sectionText(section = [], limit = 800) {
  return section.join(" ").trim().slice(0, limit);
}

function sectionItems(section = [], limit = 1200) {
  return section.filter(Boolean).join("; ").slice(0, limit);
}

function inlineSection(text, labels, nextLabels, limit) {
  const flat = text.replace(/\s+/g, " ");
  const match = flat.match(new RegExp(`\\b(?:${labels})\\b\\s*[:\\-]\\s*(.*?)(?=\\s+\\b(?:${nextLabels})\\b\\s*[:\\-]\\s*|$)`, "i"));
  return match ? match[1].trim().slice(0, limit) : "";
}

function extractCandidateFields(text) {
  const sections = parseSections(text);
  const allText = text.replace(/\s+/g, " ");
  const firstLine = firstNonContactLine(text);
  const roleKeywords = ["developer", "engineer", "manager", "analyst", "designer", "specialist", "officer", "assistant", "consultant", "accountant", "teacher", "intern"];
  let role = "";
  for (const line of (sections.other ?? []).slice(0, 8)) {
    if (line === firstLine) continue;
    const parts = line.split(/\s*(?:\||•|·)\s*|\s{2,}/);
    role = parts.find((part) => roleKeywords.some((keyword) => part.toLowerCase().includes(keyword)))?.trim() || "";
    if (role) break;
  }
  role ||= firstLine;

  const seniorityMatch = allText.match(/\b(associate|intern|internship|entry[- ]?level|junior|mid[- ]?level|intermediate|senior|lead|principal|manager|director|executive)\b/i);
  const seniority = seniorityMatch ? seniorityMatch[1].replace("-", " ").replace(/\b\w/g, (letter) => letter.toUpperCase()) : "";
  const yearsMatch = allText.match(/\b(\d+(?:\.\d+)?)\+?\s*(?:years?|yrs?)\b/i);
  const industryTerms = ["FinTech", "Banking", "Finance", "Software", "Information Technology", "Healthcare", "Education", "Marketing", "Tourism", "Retail", "Logistics", "Construction", "Manufacturing", "Telecommunications", "Insurance", "Hospitality", "Government", "E-commerce"];
  let industry = industryTerms.find((term) => new RegExp(term.replace(/[.*+?^${}()|[\]\\]/g, "\\$&"), "i").test(allText)) ?? "";
  industry ||= allText.match(/\bindustry\s*[:\-]\s*([^|]+)/i)?.[1]?.trim() || "";
  const summary = sectionText(sections.summary) || inlineSection(text, "professional\\s+summary|summary|profile|objective", "experience|education|skills|certifications", 900) || sectionText((sections.other ?? []).slice(1, 5), 900);
  const education = sectionItems(sections.education, 600) || inlineSection(text, "education|academic\\s+qualifications?", "experience|skills|summary|certifications", 600);
  const skills = sectionItems(sections.skills) || inlineSection(text, "technical\\s+skills|core\\s+skills|skills|competenc(?:y|ies)|technologies", "experience|education|summary|certifications", 1200);
  const experience = sectionItems(sections.experience, 1400) || inlineSection(text, "professional\\s+experience|work\\s+experience|experience|employment\\s+history", "education|skills|summary|certifications", 1400);

  return {
    candidateRole: role,
    candidateSeniority: seniority,
    yearsExperience: yearsMatch?.[1] ?? "",
    candidateIndustry: industry,
    education,
    candidateSkills: skills,
    summary,
    experienceBullets: experience,
  };
}

function applyCandidateFields(fields) {
  for (const [id, value] of Object.entries(fields)) $(id).value = value;
}

async function extractPdfText(file) {
  const data = new Uint8Array(await file.arrayBuffer());
  const pdf = await pdfjsLib.getDocument({ data }).promise;
  const pages = [];
  for (let pageNumber = 1; pageNumber <= pdf.numPages; pageNumber += 1) {
    const page = await pdf.getPage(pageNumber);
    const content = await page.getTextContent();
    const lines = [];
    let currentY = null;
    for (const item of content.items) {
      const text = item.str?.trim();
      if (!text) continue;
      const y = item.transform?.[5] ?? null;
      if (currentY === null || y === null || Math.abs(y - currentY) > 3) {
        lines.push(text);
        currentY = y;
      } else {
        lines[lines.length - 1] += ` ${text}`;
      }
    }
    pages.push(lines.join("\n"));
  }
  return pages.join("\n");
}

async function extractFileText(file) {
  const extension = file.name.split(".").pop().toLowerCase();
  if (extension === "txt") return file.text();
  if (extension === "pdf") return extractPdfText(file);
  if (extension === "docx") {
    if (!window.mammoth) throw new Error("DOCX reader could not be loaded. Try a PDF or TXT file.");
    const result = await window.mammoth.extractRawText({ arrayBuffer: await file.arrayBuffer() });
    return result.value;
  }
  throw new Error("Supported CV formats are PDF, DOCX and TXT.");
}

async function parseUploadedCv() {
  if (!selectedFile) return;
  extractButton.disabled = true;
  setStatus("Reading your CV in the browser…");
  try {
    const text = (await extractFileText(selectedFile)).replace(/[ \t]+/g, " ").replace(/\n{3,}/g, "\n\n").trim();
    if (!text) throw new Error("No selectable text found. This may be a scanned PDF.");
    preview.value = `${text.slice(0, 2500)}${text.length > 2500 ? "…" : ""}`;
    const fields = extractCandidateFields(text);
    applyCandidateFields(fields);
    const filled = Object.values(fields).filter(Boolean).length;
    setStatus(`Extracted ${filled} of 8 candidate fields. Please review them before scoring.`, "success");
  } catch (error) {
    setStatus(error.message || "Could not read this file.", "error");
  } finally {
    extractButton.disabled = false;
  }
}

function setSelectedFile(file) {
  if (!file) return;
  selectedFile = file;
  fileLabel.textContent = file.name;
  extractButton.disabled = false;
  setStatus("File selected. Click Extract CV details to fill the form.");
}

function buildJobText() {
  return clean(`
    Job Title: ${clean($("jobTitle").value)}.
    Required Seniority: ${clean($("jobSeniority").value)}.
    Industry: ${clean($("jobIndustry").value)}.
    Must-Have Skills: ${cleanList($("mustHaveSkills").value)}.
    Nice-to-Have Skills: ${cleanList($("niceToHaveSkills").value)}.
    Job Description: ${clean($("jobDescription").value)}.
    Responsibilities: ${cleanList($("responsibilities").value)}.
    Requirements: ${cleanList($("requirements").value)}.
  `);
}

function buildResumeText() {
  return clean(`
    Candidate Role: ${clean($("candidateRole").value)}.
    Seniority Level: ${clean($("candidateSeniority").value)}.
    Years of Experience: ${clean($("yearsExperience").value)}.
    Industry Experience: ${clean($("candidateIndustry").value)}.
    Education: ${clean($("education").value)}.
    Skills: ${cleanList($("candidateSkills").value)}.
    Professional Summary: ${clean($("summary").value)}.
    Experience Highlights: ${cleanList($("experienceBullets").value)}.
  `);
}

async function loadBrowserModel() {
  if (!modelPromise) {
    modelPromise = Promise.all([
      AutoTokenizer.from_pretrained(MODEL_ID),
      AutoModelForSequenceClassification.from_pretrained(MODEL_ID, { dtype: "fp32" }),
    ]);
  }
  return modelPromise;
}

async function scoreMatch() {
  scoreButton.disabled = true;
  scoreButton.textContent = "Loading model…";
  setStatus("The browser model is loading. The first load may take a minute.");
  result.hidden = true;
  try {
    const [tokenizer, model] = await loadBrowserModel();
    setStatus("Scoring the job and candidate profile…");
    const inputs = await tokenizer(buildJobText(), {
      text_pair: buildResumeText(),
      padding: "max_length",
      truncation: true,
      max_length: 256,
    });
    const output = await model(inputs);
    const logit = Number(output.logits.data[0]);
    const probability = 1 / (1 + Math.exp(-logit));
    const verdict = probability >= DECISION_THRESHOLD ? "Suitable match" : "Needs review";
    const percent = (probability * 100).toFixed(1);
    result.innerHTML = `
      <div class="result-top">
        <div><h3>${verdict}</h3><p>Model-estimated relevance for this job and candidate profile.</p></div>
        <div class="score-number">${percent}%</div>
      </div>
      <div class="score-bar"><span style="width:${Math.max(2, probability * 100)}%"></span></div>
      <p>Decision threshold: ${(DECISION_THRESHOLD * 100).toFixed(1)}%</p>
      <p class="result-note">This is an assistive ranking signal only. Review the original CV and job requirements before making any hiring decision.</p>
    `;
    result.hidden = false;
    setStatus("Match scored in your browser.", "success");
  } catch (error) {
    setStatus(`Could not score this profile: ${error.message || error}`, "error");
  } finally {
    scoreButton.disabled = false;
    scoreButton.innerHTML = 'Score match <span>→</span>';
  }
}

cvFile.addEventListener("change", (event) => setSelectedFile(event.target.files[0]));
extractButton.addEventListener("click", parseUploadedCv);
scoreButton.addEventListener("click", scoreMatch);
dropzone.addEventListener("dragover", (event) => { event.preventDefault(); dropzone.classList.add("dragging"); });
dropzone.addEventListener("dragleave", () => dropzone.classList.remove("dragging"));
dropzone.addEventListener("drop", (event) => {
  event.preventDefault();
  dropzone.classList.remove("dragging");
  setSelectedFile(event.dataTransfer.files[0]);
});