Delete src/app.py

src/app.py

@@ -1,354 +0,0 @@
-import streamlit as st
-import json
-import time
-from pathlib import Path
-import pandas as pd
-import numpy as np
-from sentence_transformers import SentenceTransformer
-from sklearn.feature_extraction.text import TfidfVectorizer
-
-from src.hard_filter import is_killed
-from src.score_career import compute_A, compute_keyword_max
-from src.score_skills import compute_B
-from src.score_embed import compute_C_all
-from src.availability import apply_multipliers
-from src.output import write_submission
-from src.precompute import build_candidate_text, build_jd_text
-
-# Set Page Config
-st.set_page_config(
-    page_title="Vettly Talent Intelligence Portal",
-    page_icon="💼",
-    layout="wide",
-    initial_sidebar_state="expanded"
-)
-
-# Custom HR-Themed Styling
-st.markdown("""
-<style>
-    /* Warm Ivory & Deep Indigo Theme */
-    .reportview-container {
-        background-color: #FAF8F5;
-    }
-    .sidebar .sidebar-content {
-        background-color: #1E1B4B;
-        color: #FFFFFF;
-    }
-    
-    /* Elegant Header */
-    .title-container {
-        padding: 2rem;
-        background: linear-gradient(135deg, #1E1B4B 0%, #312E81 100%);
-        color: white;
-        border-radius: 12px;
-        margin-bottom: 2rem;
-        box-shadow: 0 4px 15px rgba(0,0,0,0.05);
-    }
-    
-    /* Candidate Cards */
-    .candidate-card {
-        background-color: #FFFFFF;
-        border: 1px solid #E2E8F0;
-        border-left: 5px solid #F43F5E; /* Rose accent */
-        padding: 1.5rem;
-        border-radius: 8px;
-        margin-bottom: 1.2rem;
-        box-shadow: 0 2px 4px rgba(0,0,0,0.02);
-        transition: transform 0.2s ease, box-shadow 0.2s ease;
-    }
-    .candidate-card:hover {
-        transform: translateY(-2px);
-        box-shadow: 0 6px 12px rgba(0,0,0,0.05);
-    }
-    
-    /* Badges & Metrics */
-    .metric-badge {
-        display: inline-block;
-        padding: 0.25rem 0.6rem;
-        border-radius: 9999px;
-        font-size: 0.8rem;
-        font-weight: 600;
-        margin-right: 0.5rem;
-    }
-    .badge-rose {
-        background-color: #FFE4E6;
-        color: #E11D48;
-    }
-    .badge-indigo {
-        background-color: #E0E7FF;
-        color: #4F46E5;
-    }
-    .badge-emerald {
-        background-color: #D1FAE5;
-        color: #059669;
-    }
-    
-    /* Stats Layout */
-    .stat-box {
-        background-color: #FFFFFF;
-        border: 1px solid #E2E8F0;
-        padding: 1rem;
-        border-radius: 8px;
-        text-align: center;
-        box-shadow: 0 1px 3px rgba(0,0,0,0.01);
-    }
-    
-    /* Typography adjustments */
-    h1, h2, h3 {
-        color: #1E1B4B !important;
-        font-family: 'Inter', sans-serif;
-    }
-</style>
-""", unsafe_allow_html=True)
-
-# App Title & Welcome Banner
-st.markdown("""
-<div class="title-container">
-    <h1 style="color: white !important; margin:0; font-size: 2.2rem; font-weight:700;">Vettly Talent Portal</h1>
-    <p style="margin: 0.5rem 0 0 0; opacity: 0.9; font-size:1.1rem;">AI-Assisted Candidate Discovery, Fit Analysis, and Talent Alignment</p>
-</div>
-""", unsafe_allow_html=True)
-
-# App Title & Welcome Banner
-st.markdown("""
-<div class="title-container">
-    <h1 style="color: white !important; margin:0; font-size: 2.2rem; font-weight:700;">Vettly Talent Portal</h1>
-    <p style="margin: 0.5rem 0 0 0; opacity: 0.9; font-size:1.1rem;">AI-Assisted Candidate Discovery, Fit Analysis, and Talent Alignment</p>
-</div>
-""", unsafe_allow_html=True)
-
-# Sidebar Setup
-with st.sidebar:
-    st.markdown("<h3 style='color: white !important; margin-bottom: 1.5rem;'>📋 Upload Datasets</h3>", unsafe_allow_html=True)
-    
-    # 1. Job Description Upload
-    uploaded_jd = st.file_uploader("Upload Job Description (JSON)", type=["json"])
-    
-    # 2. Candidates Dataset Upload
-    uploaded_candidates = st.file_uploader("Upload Candidates Dataset (JSONL)", type=["jsonl"])
-    
-    st.markdown("<hr style='border-color: #312E81;'>", unsafe_allow_html=True)
-    st.markdown("<h4 style='color: white !important;'>Score Weights</h4>", unsafe_allow_html=True)
-    w_A = st.slider("Career Fit Weight (A)", 0.0, 1.0, 0.40, 0.05)
-    w_B = st.slider("Skill Trust Weight (B)", 0.0, 1.0, 0.35, 0.05)
-    w_C = st.slider("Semantic Similarity Weight (C)", 0.0, 1.0, 0.25, 0.05)
-    
-    # Check normalization
-    if abs((w_A + w_B + w_C) - 1.0) > 0.001:
-        st.warning(f"Weights sum to {w_A+w_B+w_C:.2f}. They will be normalized to 1.0 internally.")
-
-# Check if inputs are uploaded
-if not uploaded_jd or not uploaded_candidates:
-    st.info("👋 Welcome! Please upload both the **Job Description (JSON)** and **Candidates Dataset (JSONL)** in the sidebar to begin.")
-else:
-    jd_data = json.load(uploaded_jd)
-    
-    # Display JD summary info
-    col1, col2 = st.columns([1, 2])
-    with col1:
-        st.markdown("### Job Specifications")
-        st.write(f"**Target Role:** {jd_data.get('title', 'Unknown')}")
-        st.write(f"**Required Experience:** {jd_data.get('min_yoe', 5)}+ years")
-        st.write(f"**Max Budget:** {jd_data.get('budget_max_inr_lpa', 'N/A')} LPA")
-        st.write(f"**Preferred Location(s):** {', '.join(jd_data.get('preferred_locations', []))}")
-
-    with col2:
-        st.markdown("### Focus Skills & Keywords")
-        must_haves = jd_data.get("must_have_skills", [])
-        st.markdown("**Must Have Skills:**")
-        st.write(", ".join([f"`{s}`" for s in must_haves]))
-        
-        kws = jd_data.get("keywords", [])
-        st.markdown("**Target Keywords:**")
-        st.write(", ".join([f"`{k}`" for k in kws]))
-
-    # Start Button
-    if st.button("🚀 Start Talent Search & Vetting Pipeline", use_container_width=True):
-        # Normalize weights
-        total_w = w_A + w_B + w_C
-        nw_A, nw_B, nw_C = w_A/total_w, w_B/total_w, w_C/total_w
-        
-        status_box = st.empty()
-        progress_bar = st.progress(0)
-        
-        # 1. Loading & Streaming from memory buffer
-        status_box.info("Streaming uploaded candidates & fitting TF-IDF parameters...")
-        progress_bar.progress(15)
-        
-        titles = []
-        # Reset and read lines from the uploaded file buffer
-        uploaded_candidates.seek(0)
-        for line_bytes in uploaded_candidates:
-            line = line_bytes.decode("utf-8").strip()
-            if not line:
-                continue
-            cand = json.loads(line)
-            title = cand.get("profile", {}).get("current_title") or cand.get("current_title") or ""
-            titles.append(title)
-            
-        tfidf = TfidfVectorizer(max_features=30000, ngram_range=(1, 2))
-        tfidf.fit(titles)
-        del titles
-        
-        # Reset and read lines for Filtering Pass
-        uploaded_candidates.seek(0)
-
-        
-        # 2. Hard Filtering
-        status_box.info("Applying hard gatekeeper rules (Profile Completeness, Activity, Intent)...")
-        progress_bar.progress(35)
-        
-        survivors = []
-        killed_reasons = {}
-        for line_bytes in uploaded_candidates:
-            line = line_bytes.decode("utf-8").strip()
-            if not line:
-                continue
-            cand = json.loads(line)
-            killed_flag, reason = is_killed(cand, jd_data, tfidf)
-            if killed_flag:
-                # Categorize reason for display
-                category = "Other Filter"
-                if "profile_completeness_score" in reason:
-                    category = "Incomplete Profile"
-                elif "verified_email" in reason:
-                    category = "Unverified Email"
-                elif "interview_completion_rate" in reason:
-                    category = "Low Interview Completion"
-                elif "Inactive" in reason or "last_active_date" in reason:
-                    category = "Inactive > 180 Days"
-                elif "open_to_work_flag" in reason:
-                    category = "Not Open to Work"
-                elif "Zero industry overlap" in reason:
-                    category = "Industry Mismatch"
-                elif "Title similarity" in reason:
-                    category = "Role/Title Mismatch"
-                killed_reasons[category] = killed_reasons.get(category, 0) + 1
-            else:
-                survivors.append(cand)
-                    
-        # 3. Embedding Matching
-        status_box.info(f"Generating semantic candidate vectors for {len(survivors)} surviving profiles...")
-        progress_bar.progress(60)
-        
-        model = SentenceTransformer("all-MiniLM-L6-v2")
-        jd_text = build_jd_text(jd_data)
-        jd_vec = model.encode(jd_text, normalize_embeddings=True).astype("float32")
-        
-        survivor_texts = [build_candidate_text(s) for s in survivors]
-        cand_vecs = model.encode(survivor_texts, batch_size=256, normalize_embeddings=True).astype("float32")
-        
-        # 4. Scoring
-        status_box.info("Calculating comprehensive fit scores & multipliers...")
-        progress_bar.progress(85)
-        
-        C_scores = compute_C_all(jd_vec, cand_vecs)
-        C_map = {str(s.get("candidate_id") or s.get("id")): float(score) for s, score in zip(survivors, C_scores)}
-        
-        # Fit survivors TF-IDF
-        tfidf_surv = TfidfVectorizer(max_features=30000, ngram_range=(1, 2))
-        tfidf_surv.fit(survivor_texts)
-        keyword_max = compute_keyword_max(survivors, jd_data, tfidf_surv)
-        
-        raw_scored = []
-        for cand in survivors:
-            cand_id = str(cand.get("candidate_id") or cand.get("id"))
-            A_res = compute_A(cand, jd_data, tfidf_surv, keyword_max)
-            B_res = compute_B(cand, jd_data)
-            C = C_map.get(cand_id, 0.0)
-            
-            A = A_res["A"]
-            B = B_res["B"]
-            raw_score = round(nw_A * A + nw_B * B + nw_C * C, 4)
-            
-            raw_scored.append({
-                "candidate_id": cand_id,
-                "candidate": cand,
-                "A": A,
-                "B": B,
-                "C": C,
-                "raw_score": raw_score
-            })
-            
-        final_scored = apply_multipliers(raw_scored, jd_data)
-        
-        # Sort and take Top 50 for display
-        final_scored = sorted(final_scored, key=lambda x: x["final_score"], reverse=True)
-        top_candidates = final_scored[:50]
-        
-        # Clear status
-        status_box.empty()
-        progress_bar.empty()
-        
-        # Display Stats Summary Dashboard
-        st.markdown("### Talent Pipeline Summary Dashboard")
-        d_col1, d_col2, d_col3, d_col4 = st.columns(4)
-        with d_col1:
-            st.markdown(f"<div class='stat-box'><h4>Total Profiles</h4><h2 style='color:#1E1B4B;'>100,000</h2></div>", unsafe_allow_html=True)
-        with d_col2:
-            st.markdown(f"<div class='stat-box'><h4>Filtered Out</h4><h2 style='color:#E11D48;'>{100000 - len(survivors):,}</h2></div>", unsafe_allow_html=True)
-        with d_col3:
-            st.markdown(f"<div class='stat-box'><h4>Qualified Survivors</h4><h2 style='color:#059669;'>{len(survivors):,}</h2></div>", unsafe_allow_html=True)
-        with d_col4:
-            st.markdown(f"<div class='stat-box'><h4>Pruned Ratio</h4><h2 style='color:#D97706;'>{((100000 - len(survivors))/100000)*100:.2f}%</h2></div>", unsafe_allow_html=True)
-            
-        # Draw Bar chart of filtering reasons
-        st.markdown("#### Primary Reasons for Candidate Disqualification")
-        df_reasons = pd.DataFrame(list(killed_reasons.items()), columns=["Disqualification Category", "Candidate Count"])
-        st.bar_chart(df_reasons.set_index("Disqualification Category"), color="#F43F5E")
-        
-        # Output Top Candidates list in a gorgeous card design
-        st.markdown("### 🏆 Top 50 Matched Candidates")
-        
-        for rank, cand_item in enumerate(top_candidates, 1):
-            cand = cand_item["candidate"]
-            profile = cand.get("profile") or {}
-            anom_name = profile.get("anonymized_name", "Anonymous Candidate")
-            curr_title = profile.get("current_title", "Software Professional")
-            curr_company = profile.get("current_company", "N/A")
-            yoe = profile.get("years_of_experience") or profile.get("yoe") or 0.0
-            loc = profile.get("location", "Remote")
-            
-            # Scores
-            final_pct = int(cand_item["final_score"] * 100)
-            score_A_pct = int(cand_item["A"] * 100)
-            score_B_pct = int(cand_item["B"] * 100)
-            score_C_pct = int(cand_item["C"] * 100)
-            
-            # HTML Card block
-            st.markdown(f"""
-            <div class="candidate-card">
-                <div style="display:flex; justify-content:space-between; align-items:center; margin-bottom:1rem;">
-                    <div>
-                        <span class="metric-badge badge-rose" style="font-size:1rem; padding: 0.4rem 0.8rem;">Rank #{rank}</span>
-                        <strong style="font-size:1.2rem; color:#1E1B4B; margin-left: 0.5rem;">{anom_name}</strong>
-                        <span style="color:#64748B; margin-left:1rem;">{curr_title} @ {curr_company}</span>
-                    </div>
-                    <div>
-                        <span style="font-size:1.6rem; font-weight:700; color:#E11D48;">{final_pct}% Match</span>
-                    </div>
-                </div>
-                <div style="margin-bottom: 0.8rem;">
-                    <span class="metric-badge badge-indigo">💼 {yoe} Years Experience</span>
-                    <span class="metric-badge badge-indigo">📍 {loc}</span>
-                    <span class="metric-badge badge-emerald">Career Fit: {score_A_pct}%</span>
-                    <span class="metric-badge badge-emerald">Skills Trust: {score_B_pct}%</span>
-                    <span class="metric-badge badge-emerald">Semantic Sim: {score_C_pct}%</span>
-                </div>
-            </div>
-            """, unsafe_allow_html=True)
-            
-            # Details Expander
-            with st.expander(f"Inspect profile details & hiring alignment for {anom_name}"):
-                st.markdown("**Core Fit Analysis:**")
-                st.write(f"✅ Candidate has a match score of {final_pct}%. They possess {yoe} years of relevant industry experience in {profile.get('current_industry', 'tech')}. Matched locations include {loc}.")
-                
-                # Show career history
-                st.markdown("**Career History Summary:**")
-                for job in cand.get("career_history", []):
-                    st.write(f"- **{job.get('title')}** at *{job.get('company')}* ({job.get('duration_months', 0)} months) — *{job.get('description', '')[:200]}...*")
-                
-                # Show skills
-                st.markdown("**Technical Skills Inventory:**")
-                skills_list = [s.get("name") if isinstance(s, dict) else s for s in cand.get("skills", [])]
-                st.write(", ".join([f"`{s}`" for s in skills_list[:15]]))