feat: AI resume screener with scoring and ranking

quiz_generator_app.py

@@ -0,0 +1,364 @@
+import streamlit as st
+import fitz
+import os
+import requests
+import json
+import re
+import random
+
+st.set_page_config(page_title="AI Quiz Generator", page_icon="🧠", layout="wide")
+
+st.markdown("""
+<style>
+@import url('https://fonts.googleapis.com/css2?family=Plus+Jakarta+Sans:wght@300;400;500;600;700&family=JetBrains+Mono:wght@400;500&display=swap');
+html, body, [class*="css"] { font-family: 'Plus Jakarta Sans', sans-serif; }
+.main { background: #fafaf8; }
+
+.hero {
+    background: linear-gradient(135deg, #1a1a2e 0%, #16213e 50%, #0f3460 100%);
+    border-radius: 16px; padding: 32px 36px; margin-bottom: 24px; color: white;
+}
+.hero h1 { font-size: 1.9rem; font-weight: 700; margin: 0 0 6px 0; }
+.hero p { color: #94a3b8; margin: 0; font-size: 0.92rem; }
+
+.quiz-card {
+    background: white; border: 1px solid #e8e8e4;
+    border-radius: 14px; padding: 24px 28px; margin: 16px 0;
+    box-shadow: 0 2px 8px rgba(0,0,0,0.06);
+    transition: box-shadow 0.2s;
+}
+.quiz-card:hover { box-shadow: 0 4px 16px rgba(0,0,0,0.1); }
+
+.question-num {
+    font-size: 0.72rem; font-weight: 700; text-transform: uppercase;
+    letter-spacing: 0.08em; color: #94a3b8; margin-bottom: 8px;
+}
+.question-text {
+    font-size: 1.05rem; font-weight: 600; color: #1a1a2e;
+    line-height: 1.5; margin-bottom: 18px;
+}
+.difficulty-badge {
+    display: inline-block; font-size: 0.7rem; font-weight: 600;
+    padding: 3px 10px; border-radius: 20px; margin-left: 8px;
+    text-transform: uppercase; letter-spacing: 0.05em; vertical-align: middle;
+}
+.easy   { background: #dcfce7; color: #15803d; }
+.medium { background: #fef9c3; color: #854d0e; }
+.hard   { background: #fee2e2; color: #991b1b; }
+
+.option-btn {
+    display: block; width: 100%; text-align: left;
+    background: #f8f8f6; border: 2px solid #e8e8e4;
+    border-radius: 10px; padding: 12px 16px; margin: 6px 0;
+    font-size: 0.9rem; color: #374151; cursor: pointer;
+    font-family: 'Plus Jakarta Sans', sans-serif;
+    transition: all 0.15s;
+}
+.option-correct { background: #dcfce7 !important; border-color: #22c55e !important; color: #15803d !important; font-weight: 600; }
+.option-wrong   { background: #fee2e2 !important; border-color: #ef4444 !important; color: #991b1b !important; }
+.option-reveal  { background: #dbeafe !important; border-color: #3b82f6 !important; color: #1d4ed8 !important; }
+
+.explanation-box {
+    background: #f0fdf4; border: 1px solid #bbf7d0; border-left: 3px solid #22c55e;
+    border-radius: 10px; padding: 14px 18px; margin-top: 14px;
+    font-size: 0.88rem; color: #15803d; line-height: 1.6;
+}
+
+.score-display {
+    text-align: center; background: white; border: 1px solid #e8e8e4;
+    border-radius: 16px; padding: 32px; margin: 20px 0;
+    box-shadow: 0 4px 16px rgba(0,0,0,0.08);
+}
+.score-big { font-size: 4rem; font-weight: 700; color: #1a1a2e; }
+.score-label { font-size: 1rem; color: #94a3b8; margin-top: 4px; }
+
+.section-label {
+    font-size: 0.72rem; text-transform: uppercase; letter-spacing: 0.08em;
+    color: #94a3b8; font-weight: 600; margin: 20px 0 8px 0;
+}
+.stat-row { display: flex; gap: 12px; margin: 16px 0; }
+.stat-box {
+    flex: 1; background: white; border: 1px solid #e8e8e4;
+    border-radius: 10px; padding: 14px; text-align: center;
+}
+.stat-val { font-size: 1.4rem; font-weight: 700; color: #1a1a2e; }
+.stat-lbl { font-size: 0.7rem; color: #94a3b8; margin-top: 2px; }
+
+.progress-bar-bg { background: #e8e8e4; border-radius: 4px; height: 6px; margin: 8px 0; }
+.progress-bar-fill { height: 6px; border-radius: 4px; background: linear-gradient(90deg, #0f3460, #533483); }
+</style>
+""", unsafe_allow_html=True)
+
+
+# ─── Session State ────────────────────────────────────────────────────────────
+if "questions" not in st.session_state:
+    st.session_state.questions = []
+if "answers" not in st.session_state:
+    st.session_state.answers = {}
+if "revealed" not in st.session_state:
+    st.session_state.revealed = {}
+if "quiz_submitted" not in st.session_state:
+    st.session_state.quiz_submitted = False
+if "doc_title" not in st.session_state:
+    st.session_state.doc_title = ""
+
+
+# ─── Helpers ──────────────────────────────────────────────────────────────────
+def extract_pdf_text(pdf_bytes: bytes) -> str:
+    doc = fitz.open(stream=pdf_bytes, filetype="pdf")
+    text = ""
+    for page in doc:
+        text += page.get_text("text") + "\n"
+    doc.close()
+    return text.strip()
+
+
+def generate_quiz(text: str, num_questions: int, difficulty: str, topic_focus: str, api_key: str) -> list:
+    focus_instruction = f"Focus specifically on: {topic_focus}." if topic_focus else "Cover the most important concepts."
+
+    prompt = f"""You are an expert educator and assessment designer. Create a high-quality multiple choice quiz based on the document content below.
+
+Document Content:
+{text[:4000]}
+
+Requirements:
+- Generate exactly {num_questions} questions
+- Difficulty level: {difficulty}
+- {focus_instruction}
+- Each question must have exactly 4 options (A, B, C, D)
+- Only one option is correct
+- Explanations must be educational and reference the document
+
+Respond ONLY with a valid JSON array in exactly this format:
+[
+  {{
+    "question": "<clear, specific question>",
+    "options": {{
+      "A": "<option A text>",
+      "B": "<option B text>",
+      "C": "<option C text>",
+      "D": "<option D text>"
+    }},
+    "correct": "<A, B, C, or D>",
+    "explanation": "<2-3 sentence explanation of why the answer is correct, referencing the document>",
+    "difficulty": "<Easy | Medium | Hard>",
+    "topic": "<the sub-topic this question covers>"
+  }}
+]
+
+Make questions test understanding, not just memorization. Vary the question types."""
+
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+    payload = {
+        "model": "llama-3.3-70b-versatile",
+        "messages": [{"role": "user", "content": prompt}],
+        "max_tokens": 3000,
+        "temperature": 0.4,
+    }
+    r = requests.post("https://api.groq.com/openai/v1/chat/completions",
+                      headers=headers, json=payload, timeout=45)
+    r.raise_for_status()
+    raw = r.json()["choices"][0]["message"]["content"]
+    raw = re.sub(r"```json|```", "", raw).strip()
+    return json.loads(raw)
+
+
+# ─── Sidebar ──────────────────────────────────────────────────────────────────
+with st.sidebar:
+    st.markdown("## 🧠 Quiz Generator")
+    st.markdown("<div style='color:#94a3b8;font-size:0.8rem'>AI-powered assessment builder</div>", unsafe_allow_html=True)
+    st.markdown("---")
+    env_key = os.environ.get("GROQ_API_KEY", "")
+    api_key = env_key if env_key else st.text_input("🔑 Groq API Key", type="password", placeholder="gsk_...")
+    if not env_key and not api_key:
+        st.caption("Free key → [console.groq.com](https://console.groq.com)")
+    st.markdown("---")
+
+    st.markdown("<div class='section-label'>Quiz Settings</div>", unsafe_allow_html=True)
+    num_questions = st.slider("Number of Questions", min_value=3, max_value=15, value=5)
+    difficulty = st.selectbox("Difficulty", ["Mixed", "Easy", "Medium", "Hard"])
+    topic_focus = st.text_input("Topic Focus (optional)", placeholder="e.g. neural networks, photosynthesis")
+
+    st.markdown("---")
+    if st.session_state.questions:
+        total = len(st.session_state.questions)
+        answered = len(st.session_state.answers)
+        st.markdown(f"""
+<div style='font-size:0.82rem;color:#94a3b8'>
+  Progress: {answered}/{total} answered
+</div>
+<div class='progress-bar-bg'>
+  <div class='progress-bar-fill' style='width:{int(answered/total*100) if total else 0}%'></div>
+</div>
+""", unsafe_allow_html=True)
+        if st.button("🔄 Reset Quiz", use_container_width=True):
+            st.session_state.answers = {}
+            st.session_state.revealed = {}
+            st.session_state.quiz_submitted = False
+            st.rerun()
+        if st.button("🗑️ Clear & New Quiz", use_container_width=True):
+            st.session_state.questions = []
+            st.session_state.answers = {}
+            st.session_state.revealed = {}
+            st.session_state.quiz_submitted = False
+            st.session_state.doc_title = ""
+            st.rerun()
+
+
+# ─── Main UI ──────────────────────────────────────────────────────────────────
+st.markdown("""
+<div class='hero'>
+  <h1>🧠 AI Knowledge Quiz Generator</h1>
+  <p>Upload any PDF — textbook, training doc, research paper — and AI generates a complete multiple choice quiz with explanations</p>
+</div>
+""", unsafe_allow_html=True)
+
+if not api_key:
+    st.warning("👈 Add your Groq API key to get started.")
+    st.stop()
+
+# Upload section (only if no quiz yet)
+if not st.session_state.questions:
+    st.markdown("<div class='section-label'>Upload a Document</div>", unsafe_allow_html=True)
+    uploaded_file = st.file_uploader("Upload PDF", type=["pdf"], label_visibility="collapsed")
+
+    if uploaded_file:
+        st.markdown(f"<div style='font-size:0.85rem;color:#475569;padding:8px 0'>📄 {uploaded_file.name} · {round(uploaded_file.size/1024,1)}KB</div>", unsafe_allow_html=True)
+
+        if st.button(f"🧠 Generate {num_questions} Questions", type="primary", use_container_width=True):
+            with st.spinner("Reading document and generating quiz..."):
+                try:
+                    pdf_text = extract_pdf_text(uploaded_file.read())
+                    questions = generate_quiz(pdf_text, num_questions, difficulty, topic_focus, api_key)
+                    st.session_state.questions = questions
+                    st.session_state.doc_title = uploaded_file.name.replace(".pdf", "")
+                    st.session_state.answers = {}
+                    st.session_state.revealed = {}
+                    st.session_state.quiz_submitted = False
+                    st.rerun()
+                except json.JSONDecodeError:
+                    st.error("❌ AI returned unexpected format. Try again.")
+                except Exception as e:
+                    st.error(f"❌ Error: {str(e)}")
+    else:
+        st.markdown("""
+<div style='text-align:center;padding:48px 24px;border:2px dashed #e8e8e4;border-radius:14px;color:#94a3b8'>
+  <div style='font-size:2.5rem;margin-bottom:12px'>📚</div>
+  <p style='font-size:0.92rem;margin:0'>Upload a PDF to generate your quiz.<br>Works with textbooks, training docs, research papers, manuals.</p>
+</div>""", unsafe_allow_html=True)
+
+# ─── Quiz Display ──────────────────────────────────────────────────────────────
+if st.session_state.questions:
+    questions = st.session_state.questions
+    total = len(questions)
+
+    # Stats bar
+    answered = len(st.session_state.answers)
+    correct_so_far = sum(
+        1 for i, q in enumerate(questions)
+        if st.session_state.answers.get(i) == q.get("correct")
+    )
+    easy_c   = sum(1 for q in questions if q.get("difficulty") == "Easy")
+    med_c    = sum(1 for q in questions if q.get("difficulty") == "Medium")
+    hard_c   = sum(1 for q in questions if q.get("difficulty") == "Hard")
+
+    st.markdown(f"### 📝 Quiz: {st.session_state.doc_title}")
+    st.markdown(f"""
+<div class='stat-row'>
+  <div class='stat-box'><div class='stat-val'>{total}</div><div class='stat-lbl'>Questions</div></div>
+  <div class='stat-box'><div class='stat-val'>{answered}</div><div class='stat-lbl'>Answered</div></div>
+  <div class='stat-box'><div class='stat-val' style='color:#22c55e'>{easy_c}</div><div class='stat-lbl'>Easy</div></div>
+  <div class='stat-box'><div class='stat-val' style='color:#f59e0b'>{med_c}</div><div class='stat-lbl'>Medium</div></div>
+  <div class='stat-box'><div class='stat-val' style='color:#ef4444'>{hard_c}</div><div class='stat-lbl'>Hard</div></div>
+</div>
+""", unsafe_allow_html=True)
+
+    # ── Questions ──
+    for i, q in enumerate(questions):
+        diff = q.get("difficulty", "Medium")
+        diff_class = diff.lower()
+        user_answer = st.session_state.answers.get(i)
+        is_revealed = st.session_state.revealed.get(i, False)
+        is_submitted = st.session_state.quiz_submitted
+
+        with st.container():
+            st.markdown(f"""
+<div class='quiz-card'>
+  <div class='question-num'>Question {i+1} of {total} · {q.get("topic","")}
+    <span class='difficulty-badge {diff_class}'>{diff}</span>
+  </div>
+  <div class='question-text'>{q["question"]}</div>
+</div>
+""", unsafe_allow_html=True)
+
+            options = q.get("options", {})
+            correct = q.get("correct", "A")
+
+            cols = st.columns(2)
+            for j, (key, val) in enumerate(options.items()):
+                col = cols[j % 2]
+                with col:
+                    show_result = is_submitted or is_revealed
+                    if show_result:
+                        if key == correct:
+                            btn_style = "option-correct"
+                        elif key == user_answer and user_answer != correct:
+                            btn_style = "option-wrong"
+                        else:
+                            btn_style = ""
+                        st.markdown(f"<div class='option-btn {btn_style}'><b>{key}.</b> {val}</div>", unsafe_allow_html=True)
+                    else:
+                        if st.button(f"{key}. {val}", key=f"q{i}_opt_{key}", use_container_width=True):
+                            st.session_state.answers[i] = key
+                            st.rerun()
+
+            # Show selected answer indicator
+            if user_answer and not is_submitted and not is_revealed:
+                st.caption(f"✅ Selected: **{user_answer}** — {options.get(user_answer, '')}")
+
+            # Reveal/explanation
+            if is_submitted or is_revealed:
+                exp = q.get("explanation", "")
+                is_correct = user_answer == correct
+                result_text = "✅ Correct!" if is_correct else f"❌ Incorrect. Correct answer: **{correct}. {options.get(correct,'')}**"
+                st.markdown(f"""
+<div class='explanation-box'>
+  <div style='margin-bottom:6px;font-weight:600'>{result_text}</div>
+  <div>{exp}</div>
+</div>
+""", unsafe_allow_html=True)
+            elif user_answer:
+                if st.button(f"💡 Reveal Answer", key=f"reveal_{i}", use_container_width=False):
+                    st.session_state.revealed[i] = True
+                    st.rerun()
+
+            st.markdown("<div style='margin-bottom:8px'></div>", unsafe_allow_html=True)
+
+    st.markdown("---")
+
+    # Submit or Score
+    if not st.session_state.quiz_submitted:
+        col_sub, col_clear = st.columns([3, 1])
+        with col_sub:
+            if st.button("📊 Submit Quiz & See Results", type="primary", use_container_width=True,
+                         disabled=answered < total):
+                st.session_state.quiz_submitted = True
+                st.rerun()
+            if answered < total:
+                st.caption(f"Answer all {total} questions to submit. ({total - answered} remaining)")
+    else:
+        score = sum(1 for i, q in enumerate(questions)
+                    if st.session_state.answers.get(i) == q.get("correct"))
+        pct = round(score / total * 100)
+        grade = "🏆 Excellent!" if pct >= 90 else "✅ Good Job!" if pct >= 70 else "📚 Keep Studying!" if pct >= 50 else "💪 Needs Work"
+        color = "#22c55e" if pct >= 70 else "#f59e0b" if pct >= 50 else "#ef4444"
+
+        st.markdown(f"""
+<div class='score-display'>
+  <div class='score-big' style='color:{color}'>{pct}%</div>
+  <div class='score-label'>{score} out of {total} correct · {grade}</div>
+  <div style='background:#e8e8e4;border-radius:4px;height:10px;margin:16px auto;max-width:300px'>
+    <div style='height:10px;border-radius:4px;width:{pct}%;background:{color}'></div>
+  </div>
+</div>
+""", unsafe_allow_html=True)

requirements.txt

@@ -1,5 +1,4 @@
 streamlit>=1.32.0
-chromadb>=0.4.22
-sentence-transformers>=2.7.0
 requests>=2.31.0
-PyMuPDF>=1.24.0
+PyMuPDF>=1.24.0
+beautifulsoup4>=4.12.0

resume_screener_app.py

@@ -0,0 +1,283 @@
+import streamlit as st
+import fitz
+import os
+import requests
+import json
+import re
+
+st.set_page_config(page_title="AI Resume Screener", page_icon="🔍", layout="wide")
+
+st.markdown("""
+<style>
+@import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;500;600;700&family=JetBrains+Mono:wght@400;500&display=swap');
+html, body, [class*="css"] { font-family: 'Inter', sans-serif; }
+.main { background: #f8fafc; }
+
+.hero {
+    background: linear-gradient(135deg, #1e3a5f 0%, #0f2027 100%);
+    border-radius: 16px; padding: 32px 36px; margin-bottom: 24px; color: white;
+}
+.hero h1 { font-size: 1.9rem; font-weight: 700; margin: 0 0 6px 0; }
+.hero p { color: #94a3b8; margin: 0; font-size: 0.92rem; }
+
+.card {
+    background: white; border: 1px solid #e2e8f0;
+    border-radius: 12px; padding: 20px 24px; margin: 12px 0;
+    box-shadow: 0 1px 3px rgba(0,0,0,0.05);
+}
+
+.rank-1 { border-left: 4px solid #22c55e; }
+.rank-2 { border-left: 4px solid #3b82f6; }
+.rank-3 { border-left: 4px solid #f59e0b; }
+.rank-other { border-left: 4px solid #e2e8f0; }
+
+.score-badge {
+    display: inline-block; font-size: 1.4rem; font-weight: 700;
+    padding: 8px 16px; border-radius: 50px; margin-bottom: 8px;
+}
+.score-high { background: #dcfce7; color: #15803d; }
+.score-mid  { background: #dbeafe; color: #1d4ed8; }
+.score-low  { background: #fef9c3; color: #854d0e; }
+
+.candidate-name { font-size: 1.1rem; font-weight: 600; color: #1e293b; }
+.rank-label { font-size: 0.75rem; font-weight: 600; color: #64748b; text-transform: uppercase; letter-spacing: 0.05em; }
+
+.strength-tag {
+    display: inline-block; background: #dcfce7; color: #15803d;
+    border: 1px solid #bbf7d0; border-radius: 20px;
+    padding: 3px 10px; font-size: 0.78rem; margin: 2px;
+}
+.gap-tag {
+    display: inline-block; background: #fee2e2; color: #991b1b;
+    border: 1px solid #fecaca; border-radius: 20px;
+    padding: 3px 10px; font-size: 0.78rem; margin: 2px;
+}
+
+.score-bar-bg { background: #f1f5f9; border-radius: 4px; height: 8px; margin: 8px 0; }
+.score-bar-fill { height: 8px; border-radius: 4px; transition: width 0.3s; }
+
+.section-label {
+    font-size: 0.72rem; text-transform: uppercase; letter-spacing: 0.08em;
+    color: #94a3b8; font-weight: 600; margin: 20px 0 8px 0;
+}
+.stat-row { display: flex; gap: 12px; margin: 16px 0; }
+.stat-box {
+    flex: 1; background: white; border: 1px solid #e2e8f0;
+    border-radius: 10px; padding: 14px; text-align: center;
+}
+.stat-val { font-size: 1.5rem; font-weight: 700; color: #1e293b; }
+.stat-lbl { font-size: 0.72rem; color: #94a3b8; margin-top: 2px; }
+</style>
+""", unsafe_allow_html=True)
+
+
+# ─── Helpers ──────────────────────────────────────────────────────────────────
+def extract_pdf_text(pdf_bytes: bytes) -> str:
+    doc = fitz.open(stream=pdf_bytes, filetype="pdf")
+    text = ""
+    for page in doc:
+        text += page.get_text("text") + "\n"
+    doc.close()
+    return text.strip()
+
+
+def score_resume(jd_text: str, resume_text: str, candidate_name: str, api_key: str) -> dict:
+    prompt = f"""You are an expert HR recruiter and talent evaluator. Analyze the candidate's resume against the job description and provide a detailed evaluation.
+
+Job Description:
+{jd_text[:2000]}
+
+Candidate Resume ({candidate_name}):
+{resume_text[:2500]}
+
+Respond ONLY with a valid JSON object in exactly this format:
+{{
+  "score": <integer 0-100>,
+  "verdict": "<one of: Strong Match | Good Match | Partial Match | Weak Match>",
+  "summary": "<2-3 sentence overall assessment>",
+  "strengths": ["<strength 1>", "<strength 2>", "<strength 3>"],
+  "gaps": ["<gap 1>", "<gap 2>"],
+  "recommendation": "<one sentence hiring recommendation>",
+  "experience_match": <integer 0-100>,
+  "skills_match": <integer 0-100>,
+  "education_match": <integer 0-100>
+}}
+
+Be objective and specific. Base scores purely on how well the resume matches the JD requirements."""
+
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+    payload = {
+        "model": "llama-3.3-70b-versatile",
+        "messages": [{"role": "user", "content": prompt}],
+        "max_tokens": 800,
+        "temperature": 0.1,
+    }
+    r = requests.post("https://api.groq.com/openai/v1/chat/completions",
+                      headers=headers, json=payload, timeout=30)
+    r.raise_for_status()
+    raw = r.json()["choices"][0]["message"]["content"]
+    raw = re.sub(r"```json|```", "", raw).strip()
+    return json.loads(raw)
+
+
+# ─── Sidebar ──────────────────────────────────────────────────────────────────
+with st.sidebar:
+    st.markdown("## 🔍 Resume Screener")
+    st.markdown("<div style='color:#94a3b8;font-size:0.8rem'>Powered by Groq · Llama 3.3 70B</div>", unsafe_allow_html=True)
+    st.markdown("---")
+    env_key = os.environ.get("GROQ_API_KEY", "")
+    api_key = env_key if env_key else st.text_input("🔑 Groq API Key", type="password", placeholder="gsk_...")
+    if not env_key and not api_key:
+        st.caption("Free key → [console.groq.com](https://console.groq.com)")
+    st.markdown("---")
+    st.markdown("""
+<div style='font-size:0.78rem;color:#94a3b8;line-height:1.9'>
+<b>How it works</b><br>
+1. Paste the Job Description<br>
+2. Upload candidate resumes (PDF)<br>
+3. AI scores each resume 0–100<br>
+4. Candidates ranked automatically<br><br>
+<b>Scoring Dimensions</b><br>
+• Overall fit score<br>
+• Skills match %<br>
+• Experience match %<br>
+• Education match %
+</div>""", unsafe_allow_html=True)
+
+
+# ─── Main UI ──────────────────────────────────────────────────────────────────
+st.markdown("""
+<div class='hero'>
+  <h1>🔍 AI Resume Screener</h1>
+  <p>Upload a Job Description and multiple resumes — AI scores, ranks, and explains each candidate automatically</p>
+</div>
+""", unsafe_allow_html=True)
+
+col_jd, col_resumes = st.columns([1, 1], gap="large")
+
+with col_jd:
+    st.markdown("<div class='section-label'>Step 1 — Job Description</div>", unsafe_allow_html=True)
+    jd_input = st.text_area(
+        "Job Description",
+        placeholder="Paste the full job description here including role, responsibilities, required skills, and qualifications...",
+        height=320,
+        label_visibility="collapsed"
+    )
+
+with col_resumes:
+    st.markdown("<div class='section-label'>Step 2 — Upload Resumes (PDF)</div>", unsafe_allow_html=True)
+    uploaded_resumes = st.file_uploader(
+        "Upload Resumes",
+        type=["pdf"],
+        accept_multiple_files=True,
+        label_visibility="collapsed"
+    )
+    if uploaded_resumes:
+        for r in uploaded_resumes:
+            st.markdown(f"<div style='font-size:0.82rem;color:#475569;padding:4px 0'>📄 {r.name} · {round(r.size/1024,1)}KB</div>", unsafe_allow_html=True)
+
+st.markdown("")
+run_btn = st.button("🚀 Screen All Candidates", type="primary", use_container_width=True,
+                    disabled=not (jd_input and uploaded_resumes and api_key))
+
+if not api_key:
+    st.warning("👈 Add your Groq API key to get started.")
+elif not jd_input:
+    st.info("📋 Paste the job description on the left to begin.")
+elif not uploaded_resumes:
+    st.info("📂 Upload at least one resume PDF to begin.")
+
+if run_btn and jd_input and uploaded_resumes and api_key:
+    results = []
+    progress = st.progress(0, text="Screening candidates...")
+
+    for i, resume_file in enumerate(uploaded_resumes):
+        candidate_name = resume_file.name.replace(".pdf", "").replace("_", " ").replace("-", " ").title()
+        progress.progress(i / len(uploaded_resumes), text=f"Analyzing {candidate_name}...")
+
+        with st.spinner(f"Evaluating {candidate_name}..."):
+            try:
+                resume_text = extract_pdf_text(resume_file.read())
+                result = score_resume(jd_input, resume_text, candidate_name, api_key)
+                result["name"] = candidate_name
+                result["filename"] = resume_file.name
+                results.append(result)
+            except Exception as e:
+                st.error(f"❌ Error processing {candidate_name}: {str(e)}")
+
+    progress.progress(1.0, text="Screening complete!")
+
+    if results:
+        # Sort by score
+        results.sort(key=lambda x: x.get("score", 0), reverse=True)
+
+        st.markdown("---")
+        st.markdown("## 📊 Screening Results")
+
+        # Summary stats
+        avg_score = round(sum(r.get("score", 0) for r in results) / len(results))
+        top_score = results[0].get("score", 0)
+        strong = sum(1 for r in results if r.get("score", 0) >= 70)
+
+        st.markdown(f"""
+<div class='stat-row'>
+  <div class='stat-box'><div class='stat-val'>{len(results)}</div><div class='stat-lbl'>Candidates Screened</div></div>
+  <div class='stat-box'><div class='stat-val'>{top_score}</div><div class='stat-lbl'>Top Score</div></div>
+  <div class='stat-box'><div class='stat-val'>{avg_score}</div><div class='stat-lbl'>Average Score</div></div>
+  <div class='stat-box'><div class='stat-val'>{strong}</div><div class='stat-lbl'>Strong Matches</div></div>
+</div>
+""", unsafe_allow_html=True)
+
+        # Ranked results
+        for rank, result in enumerate(results, start=1):
+            score = result.get("score", 0)
+            rank_class = f"rank-{rank}" if rank <= 3 else "rank-other"
+            score_class = "score-high" if score >= 70 else "score-mid" if score >= 50 else "score-low"
+            rank_emoji = "🥇" if rank == 1 else "🥈" if rank == 2 else "🥉" if rank == 3 else f"#{rank}"
+
+            skills_w = result.get("skills_match", 0)
+            exp_w = result.get("experience_match", 0)
+            edu_w = result.get("education_match", 0)
+
+            strengths_html = "".join([f"<span class='strength-tag'>✓ {s}</span>" for s in result.get("strengths", [])])
+            gaps_html = "".join([f"<span class='gap-tag'>✗ {g}</span>" for g in result.get("gaps", [])])
+
+            with st.expander(f"{rank_emoji} {result['name']} — {score}/100 · {result.get('verdict', '')}", expanded=(rank <= 3)):
+                st.markdown(f"""
+<div class='card {rank_class}'>
+  <div style='display:flex;justify-content:space-between;align-items:flex-start;flex-wrap:wrap;gap:12px'>
+    <div>
+      <div class='rank-label'>Rank #{rank}</div>
+      <div class='candidate-name'>{result['name']}</div>
+      <div style='color:#64748b;font-size:0.82rem;margin-top:2px'>📄 {result['filename']}</div>
+    </div>
+    <div class='score-badge {score_class}'>{score} / 100</div>
+  </div>
+
+  <div style='margin:16px 0;color:#334155;font-size:0.92rem;line-height:1.7'>{result.get("summary","")}</div>
+
+  <div style='display:grid;grid-template-columns:1fr 1fr 1fr;gap:16px;margin:16px 0'>
+    <div>
+      <div style='font-size:0.75rem;color:#64748b;margin-bottom:4px'>Skills Match</div>
+      <div class='score-bar-bg'><div class='score-bar-fill' style='width:{skills_w}%;background:#3b82f6'></div></div>
+      <div style='font-size:0.78rem;font-weight:600;color:#3b82f6'>{skills_w}%</div>
+    </div>
+    <div>
+      <div style='font-size:0.75rem;color:#64748b;margin-bottom:4px'>Experience Match</div>
+      <div class='score-bar-bg'><div class='score-bar-fill' style='width:{exp_w}%;background:#22c55e'></div></div>
+      <div style='font-size:0.78rem;font-weight:600;color:#22c55e'>{exp_w}%</div>
+    </div>
+    <div>
+      <div style='font-size:0.75rem;color:#64748b;margin-bottom:4px'>Education Match</div>
+      <div class='score-bar-bg'><div class='score-bar-fill' style='width:{edu_w}%;background:#f59e0b'></div></div>
+      <div style='font-size:0.78rem;font-weight:600;color:#f59e0b'>{edu_w}%</div>
+    </div>
+  </div>
+
+  <div style='margin-bottom:10px'><div style='font-size:0.78rem;font-weight:600;color:#15803d;margin-bottom:6px'>✅ Strengths</div>{strengths_html}</div>
+  <div style='margin-bottom:14px'><div style='font-size:0.78rem;font-weight:600;color:#991b1b;margin-bottom:6px'>⚠️ Gaps</div>{gaps_html}</div>
+  <div style='background:#f8fafc;border:1px solid #e2e8f0;border-radius:8px;padding:12px;font-size:0.88rem;color:#334155'>
+    💡 <b>Recommendation:</b> {result.get("recommendation","")}
+  </div>
+</div>
+""", unsafe_allow_html=True)

sentiment_analyzer_app.py

@@ -0,0 +1,311 @@
+import streamlit as st
+import requests
+import os
+import json
+import re
+from bs4 import BeautifulSoup
+from urllib.parse import urlparse
+
+st.set_page_config(page_title="AI Sentiment Analyzer", page_icon="🌐", layout="wide")
+
+st.markdown("""
+<style>
+@import url('https://fonts.googleapis.com/css2?family=Sora:wght@300;400;500;600;700&family=JetBrains+Mono:wght@400;500&display=swap');
+html, body, [class*="css"] { font-family: 'Sora', sans-serif; }
+.main { background: #0a0a0f; }
+
+.hero {
+    background: linear-gradient(135deg, #0d0d1a 0%, #0a0a0f 100%);
+    border: 1px solid #1e1e2e; border-top: 3px solid #a78bfa;
+    border-radius: 14px; padding: 28px 32px; margin-bottom: 24px;
+}
+.hero h1 { font-size: 1.8rem; font-weight: 700; color: #f1f5f9; margin: 0 0 6px 0; }
+.hero p { color: #4b5563; font-size: 0.88rem; margin: 0; }
+
+.insight-card {
+    background: #0d0d1a; border: 1px solid #1e1e2e;
+    border-radius: 12px; padding: 20px 24px; margin: 10px 0;
+}
+
+/* Sentiment meter */
+.sentiment-positive { color: #4ade80; }
+.sentiment-negative { color: #f87171; }
+.sentiment-neutral  { color: #94a3b8; }
+.sentiment-mixed    { color: #fbbf24; }
+
+.big-sentiment {
+    font-size: 3rem; font-weight: 700; text-align: center;
+    padding: 20px; letter-spacing: -0.02em;
+}
+.sentiment-score-label {
+    text-align: center; font-size: 0.82rem; color: #4b5563;
+    font-family: 'JetBrains Mono', monospace;
+}
+
+.entity-tag {
+    display: inline-block; border-radius: 6px;
+    padding: 4px 10px; font-size: 0.78rem; margin: 3px;
+    font-family: 'JetBrains Mono', monospace;
+}
+.entity-person   { background: rgba(167,139,250,0.12); color: #a78bfa; border: 1px solid rgba(167,139,250,0.25); }
+.entity-org      { background: rgba(59,130,246,0.1);  color: #60a5fa; border: 1px solid rgba(59,130,246,0.25); }
+.entity-location { background: rgba(34,197,94,0.1);   color: #4ade80; border: 1px solid rgba(34,197,94,0.25); }
+.entity-topic    { background: rgba(251,191,36,0.1);   color: #fbbf24; border: 1px solid rgba(251,191,36,0.25); }
+.entity-product  { background: rgba(248,113,113,0.1);  color: #f87171; border: 1px solid rgba(248,113,113,0.25); }
+
+.theme-pill {
+    display: inline-block; background: #1e1e2e; border: 1px solid #2d2d3e;
+    border-radius: 20px; padding: 5px 14px; margin: 4px;
+    font-size: 0.8rem; color: #94a3b8;
+}
+
+.section-label {
+    font-size: 0.68rem; text-transform: uppercase; letter-spacing: 0.1em;
+    color: #2d2d3e; font-weight: 600; margin: 18px 0 8px 0;
+}
+
+.stat-row { display: flex; gap: 10px; margin: 16px 0; }
+.stat-box {
+    flex: 1; background: #0d0d1a; border: 1px solid #1e1e2e;
+    border-radius: 10px; padding: 14px; text-align: center;
+}
+.stat-val { font-size: 1.3rem; font-weight: 700; color: #f1f5f9; }
+.stat-lbl { font-size: 0.68rem; color: #4b5563; margin-top: 2px; }
+
+.url-chip {
+    background: #0d0d1a; border: 1px solid #1e1e2e; border-radius: 8px;
+    padding: 10px 14px; font-family: 'JetBrains Mono', monospace;
+    font-size: 0.78rem; color: #4b5563; word-break: break-all;
+    margin-bottom: 16px;
+}
+
+section[data-testid="stSidebar"] { background: #060609; border-right: 1px solid #1e1e2e; }
+</style>
+""", unsafe_allow_html=True)
+
+
+def fetch_url_text(url: str) -> tuple[str, str]:
+    headers = {"User-Agent": "Mozilla/5.0 (compatible; InsightBot/1.0)"}
+    r = requests.get(url, headers=headers, timeout=15)
+    r.raise_for_status()
+    soup = BeautifulSoup(r.text, "html.parser")
+    title = soup.title.string.strip() if soup.title else urlparse(url).netloc
+    for tag in soup(["script", "style", "nav", "footer", "header", "aside", "form"]):
+        tag.decompose()
+    text = soup.get_text(separator=" ", strip=True)
+    text = re.sub(r'\s+', ' ', text).strip()
+    return text[:4000], title
+
+
+def analyze_content(text: str, url: str, title: str, api_key: str) -> dict:
+    prompt = f"""You are an expert content analyst. Analyze the following webpage content and extract deep insights.
+
+Source URL: {url}
+Page Title: {title}
+
+Content:
+{text}
+
+Respond ONLY with a valid JSON object in exactly this format:
+{{
+  "sentiment": "<one of: Positive | Negative | Neutral | Mixed>",
+  "sentiment_score": <float between -1.0 (very negative) and 1.0 (very positive)>,
+  "sentiment_explanation": "<1-2 sentences explaining the sentiment>",
+  "one_line_summary": "<single sentence capturing the entire content>",
+  "key_themes": ["<theme 1>", "<theme 2>", "<theme 3>", "<theme 4>", "<theme 5>"],
+  "named_entities": {{
+    "persons": ["<name>"],
+    "organizations": ["<org>"],
+    "locations": ["<location>"],
+    "products": ["<product>"]
+  }},
+  "content_type": "<one of: News Article | Product Page | Review | Blog Post | Research | Social Media | Other>",
+  "target_audience": "<who this content is written for>",
+  "key_insights": ["<insight 1>", "<insight 2>", "<insight 3>"],
+  "tone": "<one of: Informative | Promotional | Critical | Analytical | Emotional | Persuasive | Neutral>",
+  "credibility_signals": ["<signal 1>", "<signal 2>"],
+  "word_count_estimate": <integer>
+}}"""
+
+    headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
+    payload = {
+        "model": "llama-3.3-70b-versatile",
+        "messages": [{"role": "user", "content": prompt}],
+        "max_tokens": 1000,
+        "temperature": 0.1,
+    }
+    r = requests.post("https://api.groq.com/openai/v1/chat/completions",
+                      headers=headers, json=payload, timeout=30)
+    r.raise_for_status()
+    raw = r.json()["choices"][0]["message"]["content"]
+    raw = re.sub(r"```json|```", "", raw).strip()
+    return json.loads(raw)
+
+
+# ─── Sidebar ──────────────────────────────────────────────────────────────────
+with st.sidebar:
+    st.markdown("## 🌐 Sentiment Analyzer")
+    st.markdown("<div style='color:#2d2d3e;font-size:0.8rem'>AI-powered content intelligence</div>", unsafe_allow_html=True)
+    st.markdown("---")
+    env_key = os.environ.get("GROQ_API_KEY", "")
+    api_key = env_key if env_key else st.text_input("🔑 Groq API Key", type="password", placeholder="gsk_...")
+    if not env_key and not api_key:
+        st.caption("Free key → [console.groq.com](https://console.groq.com)")
+    st.markdown("---")
+    st.markdown("""
+<div style='font-size:0.78rem;color:#2d2d3e;line-height:2'>
+<b style='color:#4b5563'>Extracts</b><br>
+🎭 Sentiment & Score<br>
+🏷️ Key Themes<br>
+👤 Named Entities<br>
+💡 Key Insights<br>
+🎯 Target Audience<br>
+🗣️ Content Tone<br>
+📰 Content Type
+</div>""", unsafe_allow_html=True)
+    st.markdown("---")
+    st.markdown("""
+<div style='font-size:0.78rem;color:#2d2d3e;line-height:2'>
+<b style='color:#4b5563'>Try these URLs</b><br>
+• Any news article<br>
+• Amazon product page<br>
+• Wikipedia article<br>
+• Company blog post<br>
+• G2 / Trustpilot review
+</div>""", unsafe_allow_html=True)
+
+
+# ─── Main UI ──────────────────────────────────────────────────────────────────
+st.markdown("""
+<div class='hero'>
+  <h1>🌐 AI Webpage Sentiment & Insight Analyzer</h1>
+  <p>Paste any URL — AI extracts sentiment, themes, entities, tone, and key insights in seconds</p>
+</div>
+""", unsafe_allow_html=True)
+
+if not api_key:
+    st.warning("👈 Add your Groq API key in the sidebar.")
+    st.stop()
+
+st.markdown("<div class='section-label'>Paste a URL to analyze</div>", unsafe_allow_html=True)
+col_input, col_btn = st.columns([5, 1])
+with col_input:
+    url_input = st.text_input("URL", placeholder="https://...", label_visibility="collapsed")
+with col_btn:
+    analyze_btn = st.button("Analyze ➤", type="primary", use_container_width=True)
+
+# Example URLs
+st.markdown("<div class='section-label'>Quick examples</div>", unsafe_allow_html=True)
+examples = [
+    "https://en.wikipedia.org/wiki/Artificial_intelligence",
+    "https://techcrunch.com",
+    "https://www.bbc.com/news",
+]
+cols = st.columns(len(examples))
+clicked_url = None
+for i, ex in enumerate(examples):
+    parsed = urlparse(ex)
+    label = parsed.netloc
+    if cols[i].button(f"🔗 {label}", key=f"ex_{i}", use_container_width=True):
+        clicked_url = ex
+
+final_url = clicked_url or (url_input if analyze_btn else None)
+
+if final_url:
+    with st.spinner(f"Fetching and analyzing {final_url}..."):
+        try:
+            content_text, page_title = fetch_url_text(final_url)
+            result = analyze_content(content_text, final_url, page_title, api_key)
+
+            sentiment = result.get("sentiment", "Neutral")
+            score = result.get("sentiment_score", 0)
+            sentiment_color = (
+                "#4ade80" if sentiment == "Positive" else
+                "#f87171" if sentiment == "Negative" else
+                "#fbbf24" if sentiment == "Mixed" else
+                "#94a3b8"
+            )
+            sentiment_emoji = (
+                "😊" if sentiment == "Positive" else
+                "😟" if sentiment == "Negative" else
+                "😐" if sentiment == "Neutral" else "🤔"
+            )
+            score_pct = int((score + 1) / 2 * 100)
+
+            st.markdown(f"<div class='url-chip'>🔗 {final_url}</div>", unsafe_allow_html=True)
+            st.markdown(f"### 📄 {page_title}")
+
+            # Top row
+            col_sent, col_summary = st.columns([1, 2])
+
+            with col_sent:
+                st.markdown(f"""
+<div class='insight-card' style='text-align:center'>
+  <div style='font-size:0.72rem;text-transform:uppercase;letter-spacing:0.1em;color:#4b5563;margin-bottom:8px'>Sentiment</div>
+  <div style='font-size:3.5rem'>{sentiment_emoji}</div>
+  <div style='font-size:1.6rem;font-weight:700;color:{sentiment_color};margin:4px 0'>{sentiment}</div>
+  <div style='font-family:JetBrains Mono,monospace;font-size:0.8rem;color:#4b5563'>score: {score:+.2f}</div>
+  <div style='background:#1e1e2e;border-radius:4px;height:6px;margin:10px 0'>
+    <div style='height:6px;border-radius:4px;width:{score_pct}%;background:{sentiment_color}'></div>
+  </div>
+  <div style='font-size:0.78rem;color:#4b5563;margin-top:8px'>{result.get("sentiment_explanation","")}</div>
+</div>
+""", unsafe_allow_html=True)
+
+            with col_summary:
+                entities = result.get("named_entities", {})
+                persons_html   = "".join([f"<span class='entity-tag entity-person'>👤 {e}</span>"   for e in entities.get("persons", [])[:4]])
+                orgs_html      = "".join([f"<span class='entity-tag entity-org'>🏢 {e}</span>"      for e in entities.get("organizations", [])[:4]])
+                locations_html = "".join([f"<span class='entity-tag entity-location'>📍 {e}</span>" for e in entities.get("locations", [])[:3]])
+                products_html  = "".join([f"<span class='entity-tag entity-product'>📦 {e}</span>"  for e in entities.get("products", [])[:3]])
+                entities_html  = persons_html + orgs_html + locations_html + products_html or "<span style='color:#4b5563;font-size:0.82rem'>None detected</span>"
+
+                st.markdown(f"""
+<div class='insight-card'>
+  <div style='font-size:0.72rem;text-transform:uppercase;letter-spacing:0.1em;color:#4b5563;margin-bottom:10px'>One-Line Summary</div>
+  <div style='font-size:1rem;color:#f1f5f9;font-weight:500;line-height:1.6;margin-bottom:16px'>"{result.get("one_line_summary","")}"</div>
+  <div style='display:flex;gap:16px;margin-bottom:14px'>
+    <div><span style='font-size:0.72rem;color:#4b5563'>Content Type</span><br><span style='color:#a78bfa;font-weight:600;font-size:0.88rem'>{result.get("content_type","")}</span></div>
+    <div><span style='font-size:0.72rem;color:#4b5563'>Tone</span><br><span style='color:#60a5fa;font-weight:600;font-size:0.88rem'>{result.get("tone","")}</span></div>
+    <div><span style='font-size:0.72rem;color:#4b5563'>Audience</span><br><span style='color:#4ade80;font-weight:600;font-size:0.88rem'>{result.get("target_audience","")}</span></div>
+  </div>
+  <div style='font-size:0.72rem;text-transform:uppercase;letter-spacing:0.1em;color:#4b5563;margin-bottom:8px'>Named Entities</div>
+  {entities_html}
+</div>
+""", unsafe_allow_html=True)
+
+            # Themes + Insights
+            col_themes, col_insights = st.columns(2)
+
+            with col_themes:
+                themes_html = "".join([f"<div class='theme-pill'>#{t}</div>" for t in result.get("key_themes", [])])
+                st.markdown(f"""
+<div class='insight-card'>
+  <div style='font-size:0.72rem;text-transform:uppercase;letter-spacing:0.1em;color:#4b5563;margin-bottom:12px'>🏷️ Key Themes</div>
+  {themes_html}
+</div>""", unsafe_allow_html=True)
+
+            with col_insights:
+                insights_html = "".join([f"<div style='padding:8px 0;border-bottom:1px solid #1e1e2e;font-size:0.87rem;color:#94a3b8;line-height:1.6'>→ {ins}</div>" for ins in result.get("key_insights", [])])
+                st.markdown(f"""
+<div class='insight-card'>
+  <div style='font-size:0.72rem;text-transform:uppercase;letter-spacing:0.1em;color:#4b5563;margin-bottom:12px'>💡 Key Insights</div>
+  {insights_html}
+</div>""", unsafe_allow_html=True)
+
+            # Credibility
+            cred = result.get("credibility_signals", [])
+            if cred:
+                cred_html = "".join([f"<span style='background:rgba(74,222,128,0.08);border:1px solid rgba(74,222,128,0.2);border-radius:6px;padding:4px 12px;margin:3px;display:inline-block;font-size:0.8rem;color:#4ade80'>✓ {c}</span>" for c in cred])
+                st.markdown(f"""
+<div class='insight-card'>
+  <div style='font-size:0.72rem;text-transform:uppercase;letter-spacing:0.1em;color:#4b5563;margin-bottom:10px'>🛡️ Credibility Signals</div>
+  {cred_html}
+</div>""", unsafe_allow_html=True)
+
+        except requests.exceptions.ConnectionError:
+            st.error("❌ Could not reach that URL. Make sure it's publicly accessible.")
+        except json.JSONDecodeError:
+            st.error("❌ AI returned unexpected output. Try again.")
+        except Exception as e:
+            st.error(f"❌ Error: {str(e)}")

shared_requirements.txt

@@ -0,0 +1,4 @@
+streamlit>=1.32.0
+requests>=2.31.0
+PyMuPDF>=1.24.0
+beautifulsoup4>=4.12.0

src/streamlit_app.py

@@ -1,468 +1,283 @@
 import streamlit as st
-import chromadb
-from sentence_transformers import SentenceTransformer
-import fitz  # PyMuPDF
+import fitz
 import os
 import requests
+import json
 import re
-import hashlib
 
-# ─── Page Config ──────────────────────────────────────────────────────────────
-st.set_page_config(
-    page_title="PDF RAG · Upload & Ask",
-    page_icon="📂",
-    layout="wide",
-    initial_sidebar_state="expanded"
-)
+st.set_page_config(page_title="AI Resume Screener", page_icon="🔍", layout="wide")
 
-# ─── CSS ──────────────────────────────────────────────────────────────────────
 st.markdown("""
 <style>
-@import url('https://fonts.googleapis.com/css2?family=IBM+Plex+Sans:wght@300;400;500;600&family=IBM+Plex+Mono:wght@400;500&display=swap');
-
-html, body, [class*="css"] { font-family: 'IBM Plex Sans', sans-serif; }
-.main { background-color: #0b0f1a; }
+@import url('https://fonts.googleapis.com/css2?family=Inter:wght@300;400;500;600;700&family=JetBrains+Mono:wght@400;500&display=swap');
+html, body, [class*="css"] { font-family: 'Inter', sans-serif; }
+.main { background: #f8fafc; }
 
 .hero {
-    background: linear-gradient(160deg, #0d1424 0%, #0b0f1a 100%);
-    border: 1px solid #1e2a3e;
-    border-top: 3px solid #22d3ee;
-    border-radius: 12px;
-    padding: 28px 32px;
-    margin-bottom: 24px;
+    background: linear-gradient(135deg, #1e3a5f 0%, #0f2027 100%);
+    border-radius: 16px; padding: 32px 36px; margin-bottom: 24px; color: white;
 }
-.hero h1 { font-size: 1.8rem; font-weight: 600; color: #e2e8f0; margin: 0 0 6px 0; }
-.hero p { color: #64748b; font-size: 0.95rem; margin: 0; }
+.hero h1 { font-size: 1.9rem; font-weight: 700; margin: 0 0 6px 0; }
+.hero p { color: #94a3b8; margin: 0; font-size: 0.92rem; }
 
-.phase-bar {
-    display: flex; gap: 0; margin-bottom: 28px;
-    border: 1px solid #1e2a3e; border-radius: 10px; overflow: hidden;
-}
-.phase {
-    flex: 1; padding: 10px 6px; text-align: center;
-    font-size: 0.75rem; color: #4b5563; background: #0d1117;
-    border-right: 1px solid #1e2a3e; line-height: 1.5;
-}
-.phase:last-child { border-right: none; }
-.phase.done { color: #22d3ee; background: rgba(34,211,238,0.05); }
-.phase.active { color: #f8fafc; background: rgba(34,211,238,0.1); font-weight: 600; }
-.phase-icon { font-size: 1.1rem; display: block; margin-bottom: 2px; }
-
-.pdf-card {
-    background: #0d1424;
-    border: 1px solid #1e2a3e;
-    border-radius: 10px;
-    padding: 14px 16px;
-    margin: 8px 0;
-    display: flex;
-    align-items: center;
-    justify-content: space-between;
-}
-.pdf-name { font-size: 0.85rem; color: #e2e8f0; font-weight: 500; }
-.pdf-meta { font-family: 'IBM Plex Mono', monospace; font-size: 0.72rem; color: #475569; margin-top: 3px; }
-.pdf-badge {
-    font-size: 0.72rem; font-family: 'IBM Plex Mono', monospace;
-    background: rgba(34,211,238,0.1); color: #22d3ee;
-    border: 1px solid rgba(34,211,238,0.25); padding: 3px 10px; border-radius: 20px;
+.card {
+    background: white; border: 1px solid #e2e8f0;
+    border-radius: 12px; padding: 20px 24px; margin: 12px 0;
+    box-shadow: 0 1px 3px rgba(0,0,0,0.05);
 }
 
-.answer-box {
-    background: #0d1424;
-    border: 1px solid #1e3a4a;
-    border-left: 3px solid #22d3ee;
-    border-radius: 10px;
-    padding: 22px 24px;
-    color: #e2e8f0;
-    line-height: 1.75;
-    font-size: 0.96rem;
-    margin: 12px 0 20px 0;
-}
+.rank-1 { border-left: 4px solid #22c55e; }
+.rank-2 { border-left: 4px solid #3b82f6; }
+.rank-3 { border-left: 4px solid #f59e0b; }
+.rank-other { border-left: 4px solid #e2e8f0; }
 
-.chunk-card {
-    background: #0d1117;
-    border: 1px solid #1e2a3e;
-    border-radius: 9px;
-    padding: 14px 18px;
-    margin: 8px 0;
+.score-badge {
+    display: inline-block; font-size: 1.4rem; font-weight: 700;
+    padding: 8px 16px; border-radius: 50px; margin-bottom: 8px;
 }
-.chunk-top {
-    display: flex; justify-content: space-between;
-    align-items: center; margin-bottom: 8px;
+.score-high { background: #dcfce7; color: #15803d; }
+.score-mid  { background: #dbeafe; color: #1d4ed8; }
+.score-low  { background: #fef9c3; color: #854d0e; }
+
+.candidate-name { font-size: 1.1rem; font-weight: 600; color: #1e293b; }
+.rank-label { font-size: 0.75rem; font-weight: 600; color: #64748b; text-transform: uppercase; letter-spacing: 0.05em; }
+
+.strength-tag {
+    display: inline-block; background: #dcfce7; color: #15803d;
+    border: 1px solid #bbf7d0; border-radius: 20px;
+    padding: 3px 10px; font-size: 0.78rem; margin: 2px;
 }
-.chunk-source { font-size: 0.77rem; font-weight: 600; color: #22d3ee; text-transform: uppercase; letter-spacing: 0.05em; }
-.chunk-page { font-family: 'IBM Plex Mono', monospace; font-size: 0.72rem; color: #475569; }
-.score-bar-wrap { display: flex; align-items: center; gap: 8px; }
-.score-bar {
-    height: 4px; border-radius: 2px; background: #1e2a3e; width: 80px; overflow: hidden;
+.gap-tag {
+    display: inline-block; background: #fee2e2; color: #991b1b;
+    border: 1px solid #fecaca; border-radius: 20px;
+    padding: 3px 10px; font-size: 0.78rem; margin: 2px;
 }
-.score-fill { height: 100%; border-radius: 2px; background: #22d3ee; }
-.score-num { font-family: 'IBM Plex Mono', monospace; font-size: 0.72rem; color: #22d3ee; }
-.chunk-text { font-size: 0.86rem; color: #94a3b8; line-height: 1.65; }
 
-.stat-row { display: flex; gap: 10px; margin: 16px 0; }
-.stat-box {
-    flex: 1; background: #0d1424; border: 1px solid #1e2a3e;
-    border-radius: 8px; padding: 12px; text-align: center;
-}
-.stat-val { font-size: 1.35rem; font-weight: 600; color: #22d3ee; }
-.stat-lbl { font-size: 0.7rem; color: #475569; margin-top: 2px; }
+.score-bar-bg { background: #f1f5f9; border-radius: 4px; height: 8px; margin: 8px 0; }
+.score-bar-fill { height: 8px; border-radius: 4px; transition: width 0.3s; }
 
 .section-label {
-    font-size: 0.7rem; text-transform: uppercase; letter-spacing: 0.1em;
-    color: #374151; font-weight: 600; margin: 18px 0 8px 0;
+    font-size: 0.72rem; text-transform: uppercase; letter-spacing: 0.08em;
+    color: #94a3b8; font-weight: 600; margin: 20px 0 8px 0;
 }
-
-section[data-testid="stSidebar"] {
-    background-color: #080c14; border-right: 1px solid #131c2e;
-}
-
-.empty-state {
-    text-align: center; padding: 48px 24px;
-    border: 2px dashed #1e2a3e; border-radius: 12px; color: #374151;
+.stat-row { display: flex; gap: 12px; margin: 16px 0; }
+.stat-box {
+    flex: 1; background: white; border: 1px solid #e2e8f0;
+    border-radius: 10px; padding: 14px; text-align: center;
 }
-.empty-state .icon { font-size: 2.5rem; margin-bottom: 12px; }
-.empty-state p { font-size: 0.9rem; line-height: 1.6; }
+.stat-val { font-size: 1.5rem; font-weight: 700; color: #1e293b; }
+.stat-lbl { font-size: 0.72rem; color: #94a3b8; margin-top: 2px; }
 </style>
 """, unsafe_allow_html=True)
 
 
-# ─── Session State ────────────────────────────────────────────────────────────
-if "indexed_files" not in st.session_state:
-    st.session_state.indexed_files = {}   # filename → {chunks, pages, size}
-if "chroma_collection" not in st.session_state:
-    st.session_state.chroma_collection = None
-if "chroma_client" not in st.session_state:
-    st.session_state.chroma_client = None
-if "total_chunks" not in st.session_state:
-    st.session_state.total_chunks = 0
-
-
-# ─── Load embedding model (cached globally) ───────────────────────────────────
-@st.cache_resource(show_spinner=False)
-def load_embed_model():
-    return SentenceTransformer('all-MiniLM-L6-v2')
-
-
-# ─── PDF Extraction ───────────────────────────────────────────────────────────
-def extract_text_from_pdf(pdf_bytes: bytes) -> list[dict]:
-    """Returns list of {page, text} dicts."""
+# ─── Helpers ──────────────────────────────────────────────────────────────────
+def extract_pdf_text(pdf_bytes: bytes) -> str:
     doc = fitz.open(stream=pdf_bytes, filetype="pdf")
-    pages = []
-    for page_num, page in enumerate(doc, start=1):
-        text = page.get_text("text").strip()
-        if text:
-            pages.append({"page": page_num, "text": text})
+    text = ""
+    for page in doc:
+        text += page.get_text("text") + "\n"
     doc.close()
-    return pages
-
-
-# ─── Chunking ─────────────────────────────────────────────────────────────────
-def chunk_text(pages: list[dict], chunk_size: int = 400, overlap: int = 60) -> list[dict]:
-    """Splits page text into overlapping word-based chunks."""
-    chunks = []
-    for p in pages:
-        words = p["text"].split()
-        start = 0
-        while start < len(words):
-            end = start + chunk_size
-            chunk_words = words[start:end]
-            chunk_text_str = " ".join(chunk_words).strip()
-            if len(chunk_text_str) > 60:
-                chunks.append({"text": chunk_text_str, "page": p["page"]})
-            start += chunk_size - overlap
-    return chunks
-
-
-# ─── Index PDF into ChromaDB ──────────────────────────────────────────────────
-def index_pdf(filename: str, pdf_bytes: bytes, embed_model):
-    # Init or reuse ChromaDB
-    if st.session_state.chroma_client is None:
-        st.session_state.chroma_client = chromadb.Client()
-        st.session_state.chroma_collection = st.session_state.chroma_client.get_or_create_collection(
-            name="pdf_rag", metadata={"hnsw:space": "cosine"}
-        )
-
-    collection = st.session_state.chroma_collection
-
-    # Extract & chunk
-    pages = extract_text_from_pdf(pdf_bytes)
-    chunks = chunk_text(pages)
-
-    if not chunks:
-        return 0, 0
-
-    # Embed & add
-    texts = [c["text"] for c in chunks]
-    embeddings = embed_model.encode(texts, batch_size=32, show_progress_bar=False).tolist()
-
-    ids, docs, metas, embeds = [], [], [], []
-    for i, (chunk, emb) in enumerate(zip(chunks, embeddings)):
-        chunk_id = f"{hashlib.md5(filename.encode()).hexdigest()[:8]}_chunk_{i}"
-        ids.append(chunk_id)
-        docs.append(chunk["text"])
-        metas.append({"filename": filename, "page": chunk["page"]})
-        embeds.append(emb)
-
-    collection.add(ids=ids, embeddings=embeds, documents=docs, metadatas=metas)
-
-    st.session_state.indexed_files[filename] = {
-        "chunks": len(chunks),
-        "pages": len(pages),
-        "size_kb": round(len(pdf_bytes) / 1024, 1)
-    }
-    st.session_state.total_chunks += len(chunks)
-    return len(chunks), len(pages)
+    return text.strip()
 
 
-# ─── RAG Query ────────────────────────────────────────────────────────────────
-def rag_query(question: str, embed_model, top_k: int, api_key: str):
-    collection = st.session_state.chroma_collection
-    q_emb = embed_model.encode(question).tolist()
-    results = collection.query(query_embeddings=[q_emb], n_results=top_k)
+def score_resume(jd_text: str, resume_text: str, candidate_name: str, api_key: str) -> dict:
+    prompt = f"""You are an expert HR recruiter and talent evaluator. Analyze the candidate's resume against the job description and provide a detailed evaluation.
 
-    chunks = []
-    for i in range(len(results["documents"][0])):
-        distance = results["distances"][0][i]
-        chunks.append({
-            "text": results["documents"][0][i],
-            "filename": results["metadatas"][0][i]["filename"],
-            "page": results["metadatas"][0][i]["page"],
-            "relevance": round((1 - distance) * 100, 1),
-        })
+Job Description:
+{jd_text[:2000]}
 
-    context = "\n\n".join([
-        f"[Source: {c['filename']}, Page {c['page']}]\n{c['text']}" for c in chunks
-    ])
+Candidate Resume ({candidate_name}):
+{resume_text[:2500]}
 
-    prompt = f"""You are a helpful assistant. Answer the user's question using ONLY the document context provided below. Be concise and clear. Always mention the source filename and page number when referencing specific information. If the answer cannot be found in the provided context, say "I couldn't find that information in the uploaded documents."
+Respond ONLY with a valid JSON object in exactly this format:
+{{
+  "score": <integer 0-100>,
+  "verdict": "<one of: Strong Match | Good Match | Partial Match | Weak Match>",
+  "summary": "<2-3 sentence overall assessment>",
+  "strengths": ["<strength 1>", "<strength 2>", "<strength 3>"],
+  "gaps": ["<gap 1>", "<gap 2>"],
+  "recommendation": "<one sentence hiring recommendation>",
+  "experience_match": <integer 0-100>,
+  "skills_match": <integer 0-100>,
+  "education_match": <integer 0-100>
+}}
 
-Document Context:
-{context}
-
-Question: {question}
-
-Answer:"""
+Be objective and specific. Base scores purely on how well the resume matches the JD requirements."""
 
     headers = {"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}
     payload = {
         "model": "llama-3.3-70b-versatile",
         "messages": [{"role": "user", "content": prompt}],
-        "max_tokens": 600,
-        "temperature": 0.2,
+        "max_tokens": 800,
+        "temperature": 0.1,
     }
-    r = requests.post("https://api.groq.com/openai/v1/chat/completions", headers=headers, json=payload, timeout=30)
+    r = requests.post("https://api.groq.com/openai/v1/chat/completions",
+                      headers=headers, json=payload, timeout=30)
     r.raise_for_status()
-    answer = r.json()["choices"][0]["message"]["content"]
-    return answer, chunks
-
-
-# ─── Determine current phase ──────────────────────────────────────────────────
-has_docs = len(st.session_state.indexed_files) > 0
-phase = 1 if not has_docs else 2
+    raw = r.json()["choices"][0]["message"]["content"]
+    raw = re.sub(r"```json|```", "", raw).strip()
+    return json.loads(raw)
 
 
 # ─── Sidebar ──────────────────────────────────────────────────────────────────
 with st.sidebar:
-    st.markdown("## 📂 PDF RAG Demo")
-    st.markdown("<div style='color:#374151;font-size:0.8rem'>Upload → Extract → Index → Ask</div>", unsafe_allow_html=True)
+    st.markdown("## 🔍 Resume Screener")
+    st.markdown("<div style='color:#94a3b8;font-size:0.8rem'>Powered by Groq · Llama 3.3 70B</div>", unsafe_allow_html=True)
     st.markdown("---")
-
     env_key = os.environ.get("GROQ_API_KEY", "")
-    if env_key:
-        api_key = env_key
-        st.success("✅ Groq key loaded from secrets")
-    else:
-        api_key = st.text_input("🔑 Groq API Key", type="password", placeholder="gsk_...", help="Free at console.groq.com")
-        if not api_key:
-            st.caption("Get free key → [console.groq.com](https://console.groq.com)")
-
-    st.markdown("---")
-    st.markdown("<div class='section-label'>Indexed Documents</div>", unsafe_allow_html=True)
-
-    if st.session_state.indexed_files:
-        for fname, info in st.session_state.indexed_files.items():
-            st.markdown(f"""
-<div style='padding:6px 0;border-bottom:1px solid #131c2e'>
-  <div style='font-size:0.8rem;color:#e2e8f0'>📄 {fname}</div>
-  <div style='font-size:0.72rem;color:#475569;font-family:IBM Plex Mono,monospace'>
-    {info["pages"]}p · {info["chunks"]} chunks · {info["size_kb"]}KB
-  </div>
-</div>""", unsafe_allow_html=True)
-
-        st.markdown("---")
-        if st.button("🗑️ Clear all & reset", use_container_width=True):
-            for key in ["indexed_files", "chroma_collection", "chroma_client", "total_chunks"]:
-                del st.session_state[key]
-            st.rerun()
-    else:
-        st.markdown("<div style='color:#374151;font-size:0.82rem'>No documents indexed yet.</div>", unsafe_allow_html=True)
-
+    api_key = env_key if env_key else st.text_input("🔑 Groq API Key", type="password", placeholder="gsk_...")
+    if not env_key and not api_key:
+        st.caption("Free key → [console.groq.com](https://console.groq.com)")
     st.markdown("---")
     st.markdown("""
-<div style='font-size:0.77rem;color:#374151;line-height:1.9'>
-<b style='color:#4b5563'>Stack</b><br>
-📄 PDF parsing: PyMuPDF<br>
-✂️ Chunking: word-overlap (400w)<br>
-🔢 Embeddings: all-MiniLM-L6-v2<br>
-🗄️ Vector DB: ChromaDB in-memory<br>
-🧠 LLM: Groq · Llama 3.3 70B<br>
-🌐 Hosting: HuggingFace Spaces
-</div>
-""", unsafe_allow_html=True)
+<div style='font-size:0.78rem;color:#94a3b8;line-height:1.9'>
+<b>How it works</b><br>
+1. Paste the Job Description<br>
+2. Upload candidate resumes (PDF)<br>
+3. AI scores each resume 0–100<br>
+4. Candidates ranked automatically<br><br>
+<b>Scoring Dimensions</b><br>
+• Overall fit score<br>
+• Skills match %<br>
+• Experience match %<br>
+• Education match %
+</div>""", unsafe_allow_html=True)
 
 
-# ─── Hero ─────────────────────────────────────────────────────────────────────
+# ─── Main UI ──────────────────────────────────────────────────────────────────
 st.markdown("""
 <div class='hero'>
-  <h1>📂 PDF RAG — Upload & Ask</h1>
-  <p>Upload any PDF documents · They get extracted, chunked, embedded, and indexed · Then ask questions across all of them</p>
-</div>
-""", unsafe_allow_html=True)
-
-# Phase bar
-st.markdown(f"""
-<div class='phase-bar'>
-  <div class='phase {"done" if phase > 1 else "active"}'>
-    <span class='phase-icon'>📤</span>Upload PDFs
-  </div>
-  <div class='phase {"active" if phase == 1 else "done"}'>
-    <span class='phase-icon'>📑</span>Extract Text
-  </div>
-  <div class='phase {"active" if phase == 1 else "done"}'>
-    <span class='phase-icon'>✂️</span>Chunk
-  </div>
-  <div class='phase {"active" if phase == 1 else "done"}'>
-    <span class='phase-icon'>🔢</span>Embed
-  </div>
-  <div class='phase {"active" if phase == 1 else "done"}'>
-    <span class='phase-icon'>🗄️</span>Index
-  </div>
-  <div class='phase {"active" if phase == 2 else ""}'>
-    <span class='phase-icon'>💬</span>Ask Questions
-  </div>
+  <h1>🔍 AI Resume Screener</h1>
+  <p>Upload a Job Description and multiple resumes — AI scores, ranks, and explains each candidate automatically</p>
 </div>
 """, unsafe_allow_html=True)
 
-# ─── Load model ───────────────────────────────────────────────────────────────
-with st.spinner("⚙️ Loading embedding model..."):
-    embed_model = load_embed_model()
-
-
-# ════════════════════════════════════════════════════════════
-# PHASE 1 — Upload & Index
-# ════════════════════════════════════════════════════════════
-st.markdown("<div class='section-label'>Step 1 — Upload PDF Documents</div>", unsafe_allow_html=True)
-
-uploaded_files = st.file_uploader(
-    "Drop your PDF files here",
-    type=["pdf"],
-    accept_multiple_files=True,
-    label_visibility="collapsed"
-)
-
-if uploaded_files:
-    new_files = [f for f in uploaded_files if f.name not in st.session_state.indexed_files]
-
-    if new_files:
-        st.markdown(f"**{len(new_files)} new file(s) ready to index:**")
-        for f in new_files:
-            st.markdown(f"<div class='pdf-card'><div><div class='pdf-name'>📄 {f.name}</div><div class='pdf-meta'>{round(f.size/1024,1)} KB</div></div><div class='pdf-badge'>ready</div></div>", unsafe_allow_html=True)
-
-        if st.button(f"⚡ Extract & Index {len(new_files)} PDF(s)", type="primary", use_container_width=True):
-            progress = st.progress(0, text="Starting...")
-            for idx, f in enumerate(new_files):
-                progress.progress((idx) / len(new_files), text=f"Processing: {f.name}")
-                pdf_bytes = f.read()
-
-                with st.spinner(f"Extracting & indexing **{f.name}**..."):
-                    n_chunks, n_pages = index_pdf(f.name, pdf_bytes, embed_model)
-
-                st.success(f"✅ **{f.name}** → {n_pages} pages · {n_chunks} chunks indexed")
-
-            progress.progress(1.0, text="Done!")
-            st.balloons()
-            st.rerun()
+col_jd, col_resumes = st.columns([1, 1], gap="large")
+
+with col_jd:
+    st.markdown("<div class='section-label'>Step 1 — Job Description</div>", unsafe_allow_html=True)
+    jd_input = st.text_area(
+        "Job Description",
+        placeholder="Paste the full job description here including role, responsibilities, required skills, and qualifications...",
+        height=320,
+        label_visibility="collapsed"
+    )
+
+with col_resumes:
+    st.markdown("<div class='section-label'>Step 2 — Upload Resumes (PDF)</div>", unsafe_allow_html=True)
+    uploaded_resumes = st.file_uploader(
+        "Upload Resumes",
+        type=["pdf"],
+        accept_multiple_files=True,
+        label_visibility="collapsed"
+    )
+    if uploaded_resumes:
+        for r in uploaded_resumes:
+            st.markdown(f"<div style='font-size:0.82rem;color:#475569;padding:4px 0'>📄 {r.name} · {round(r.size/1024,1)}KB</div>", unsafe_allow_html=True)
+
+st.markdown("")
+run_btn = st.button("🚀 Screen All Candidates", type="primary", use_container_width=True,
+                    disabled=not (jd_input and uploaded_resumes and api_key))
+
+if not api_key:
+    st.warning("👈 Add your Groq API key to get started.")
+elif not jd_input:
+    st.info("📋 Paste the job description on the left to begin.")
+elif not uploaded_resumes:
+    st.info("📂 Upload at least one resume PDF to begin.")
+
+if run_btn and jd_input and uploaded_resumes and api_key:
+    results = []
+    progress = st.progress(0, text="Screening candidates...")
+
+    for i, resume_file in enumerate(uploaded_resumes):
+        candidate_name = resume_file.name.replace(".pdf", "").replace("_", " ").replace("-", " ").title()
+        progress.progress(i / len(uploaded_resumes), text=f"Analyzing {candidate_name}...")
+
+        with st.spinner(f"Evaluating {candidate_name}..."):
+            try:
+                resume_text = extract_pdf_text(resume_file.read())
+                result = score_resume(jd_input, resume_text, candidate_name, api_key)
+                result["name"] = candidate_name
+                result["filename"] = resume_file.name
+                results.append(result)
+            except Exception as e:
+                st.error(f"❌ Error processing {candidate_name}: {str(e)}")
 
-    else:
-        st.info("All uploaded files are already indexed. Upload new files or ask questions below.")
+    progress.progress(1.0, text="Screening complete!")
 
-elif not has_docs:
-    st.markdown("""
-<div class='empty-state'>
-  <div class='icon'>📂</div>
-  <p><b style='color:#94a3b8'>No documents uploaded yet</b><br>
-  Upload one or more PDF files above to get started.<br>
-  Any topic works — reports, manuals, research papers, policies.</p>
-</div>
-""", unsafe_allow_html=True)
+    if results:
+        # Sort by score
+        results.sort(key=lambda x: x.get("score", 0), reverse=True)
 
+        st.markdown("---")
+        st.markdown("## 📊 Screening Results")
 
-# ════════════════════════════════════════════════════════════
-# PHASE 2 — Stats & Query
-# ════════════════════════════════════════════════════════════
-if has_docs:
-    total_pages = sum(v["pages"] for v in st.session_state.indexed_files.values())
+        # Summary stats
+        avg_score = round(sum(r.get("score", 0) for r in results) / len(results))
+        top_score = results[0].get("score", 0)
+        strong = sum(1 for r in results if r.get("score", 0) >= 70)
 
-    st.markdown("<div class='section-label' style='margin-top:24px'>Index Summary</div>", unsafe_allow_html=True)
-    st.markdown(f"""
+        st.markdown(f"""
 <div class='stat-row'>
-  <div class='stat-box'><div class='stat-val'>{len(st.session_state.indexed_files)}</div><div class='stat-lbl'>Documents</div></div>
-  <div class='stat-box'><div class='stat-val'>{total_pages}</div><div class='stat-lbl'>Pages Parsed</div></div>
-  <div class='stat-box'><div class='stat-val'>{st.session_state.total_chunks}</div><div class='stat-lbl'>Chunks Indexed</div></div>
-  <div class='stat-box'><div class='stat-val'>384</div><div class='stat-lbl'>Embedding Dims</div></div>
+  <div class='stat-box'><div class='stat-val'>{len(results)}</div><div class='stat-lbl'>Candidates Screened</div></div>
+  <div class='stat-box'><div class='stat-val'>{top_score}</div><div class='stat-lbl'>Top Score</div></div>
+  <div class='stat-box'><div class='stat-val'>{avg_score}</div><div class='stat-lbl'>Average Score</div></div>
+  <div class='stat-box'><div class='stat-val'>{strong}</div><div class='stat-lbl'>Strong Matches</div></div>
 </div>
 """, unsafe_allow_html=True)
 
-    if not api_key:
-        st.warning("👈 Enter your Groq API key in the sidebar to start asking questions.")
-        st.stop()
-
-    st.markdown("---")
-    st.markdown("<div class='section-label'>Step 2 — Ask a Question</div>", unsafe_allow_html=True)
+        # Ranked results
+        for rank, result in enumerate(results, start=1):
+            score = result.get("score", 0)
+            rank_class = f"rank-{rank}" if rank <= 3 else "rank-other"
+            score_class = "score-high" if score >= 70 else "score-mid" if score >= 50 else "score-low"
+            rank_emoji = "🥇" if rank == 1 else "🥈" if rank == 2 else "🥉" if rank == 3 else f"#{rank}"
 
-    col1, col2 = st.columns([5, 1])
-    with col1:
-        question = st.text_input("", placeholder="What does the document say about...?", label_visibility="collapsed")
-    with col2:
-        top_k = st.selectbox("Top K", [2, 3, 4, 5], index=1, help="Number of chunks to retrieve")
+            skills_w = result.get("skills_match", 0)
+            exp_w = result.get("experience_match", 0)
+            edu_w = result.get("education_match", 0)
 
-    ask_btn = st.button("🔍 Search & Answer", type="primary", use_container_width=True)
+            strengths_html = "".join([f"<span class='strength-tag'>✓ {s}</span>" for s in result.get("strengths", [])])
+            gaps_html = "".join([f"<span class='gap-tag'>✗ {g}</span>" for g in result.get("gaps", [])])
 
-    if ask_btn and question:
-        with st.spinner("🔍 Searching index and generating answer..."):
-            try:
-                answer, chunks = rag_query(question, embed_model, top_k, api_key)
-
-                st.markdown(f"<div class='section-label'>Answer</div>", unsafe_allow_html=True)
-                st.markdown(f"<div class='answer-box'>{answer}</div>", unsafe_allow_html=True)
+            with st.expander(f"{rank_emoji} {result['name']} — {score}/100 · {result.get('verdict', '')}", expanded=(rank <= 3)):
+                st.markdown(f"""
+<div class='card {rank_class}'>
+  <div style='display:flex;justify-content:space-between;align-items:flex-start;flex-wrap:wrap;gap:12px'>
+    <div>
+      <div class='rank-label'>Rank #{rank}</div>
+      <div class='candidate-name'>{result['name']}</div>
+      <div style='color:#64748b;font-size:0.82rem;margin-top:2px'>📄 {result['filename']}</div>
+    </div>
+    <div class='score-badge {score_class}'>{score} / 100</div>
+  </div>
 
-                st.markdown("<div class='section-label'>Retrieved Chunks (context sent to LLM)</div>", unsafe_allow_html=True)
+  <div style='margin:16px 0;color:#334155;font-size:0.92rem;line-height:1.7'>{result.get("summary","")}</div>
 
-                for i, chunk in enumerate(chunks):
-                    bar_width = int(chunk['relevance'])
-                    st.markdown(f"""
-<div class='chunk-card'>
-  <div class='chunk-top'>
+  <div style='display:grid;grid-template-columns:1fr 1fr 1fr;gap:16px;margin:16px 0'>
     <div>
-      <div class='chunk-source'>📄 {chunk['filename']}</div>
-      <div class='chunk-page'>Page {chunk['page']}</div>
+      <div style='font-size:0.75rem;color:#64748b;margin-bottom:4px'>Skills Match</div>
+      <div class='score-bar-bg'><div class='score-bar-fill' style='width:{skills_w}%;background:#3b82f6'></div></div>
+      <div style='font-size:0.78rem;font-weight:600;color:#3b82f6'>{skills_w}%</div>
     </div>
-    <div class='score-bar-wrap'>
-      <div class='score-bar'><div class='score-fill' style='width:{bar_width}%'></div></div>
-      <div class='score-num'>{chunk['relevance']}%</div>
+    <div>
+      <div style='font-size:0.75rem;color:#64748b;margin-bottom:4px'>Experience Match</div>
+      <div class='score-bar-bg'><div class='score-bar-fill' style='width:{exp_w}%;background:#22c55e'></div></div>
+      <div style='font-size:0.78rem;font-weight:600;color:#22c55e'>{exp_w}%</div>
+    </div>
+    <div>
+      <div style='font-size:0.75rem;color:#64748b;margin-bottom:4px'>Education Match</div>
+      <div class='score-bar-bg'><div class='score-bar-fill' style='width:{edu_w}%;background:#f59e0b'></div></div>
+      <div style='font-size:0.78rem;font-weight:600;color:#f59e0b'>{edu_w}%</div>
     </div>
   </div>
-  <div class='chunk-text'>{chunk['text']}</div>
+
+  <div style='margin-bottom:10px'><div style='font-size:0.78rem;font-weight:600;color:#15803d;margin-bottom:6px'>✅ Strengths</div>{strengths_html}</div>
+  <div style='margin-bottom:14px'><div style='font-size:0.78rem;font-weight:600;color:#991b1b;margin-bottom:6px'>⚠️ Gaps</div>{gaps_html}</div>
+  <div style='background:#f8fafc;border:1px solid #e2e8f0;border-radius:8px;padding:12px;font-size:0.88rem;color:#334155'>
+    💡 <b>Recommendation:</b> {result.get("recommendation","")}
+  </div>
 </div>
 """, unsafe_allow_html=True)
-
-            except requests.HTTPError as e:
-                if e.response.status_code == 401:
-                    st.error("❌ Invalid Groq API key.")
-                else:
-                    st.error(f"❌ API error: {str(e)}")
-            except Exception as e:
-                st.error(f"❌ Error: {str(e)}")
-
-    elif ask_btn and not question:
-        st.warning("Please enter a question.")