ss

app.py

@@ -47,14 +47,8 @@ st.set_page_config(
 with st.sidebar:
     st.title("⚙️ Configuration")
     
-    # Ranking weights
-    st.subheader("Ranking Weights")
-    semantic_weight = st.slider("Semantic Similarity Weight", 0.0, 1.0, 0.7, 0.1)
-    keyword_weight = 1.0 - semantic_weight
-    st.write(f"Keyword Weight: {keyword_weight:.1f}")
-    
     # Advanced options
-    st.subheader("Advanced Options")
+    st.subheader("Display Options")
     top_k = st.selectbox("Number of results to display", options=[1, 2, 3, 4, 5], index=4)
     
     # LLM Settings
@@ -62,18 +56,16 @@ with st.sidebar:
     st.info("💡 Intent analysis using Qwen3-1.7B is always enabled")
     
     st.markdown("---")
-    st.markdown("### 🤖 Advanced Pipeline")
-    st.markdown("- **Stage 1**: FAISS Recall (Top 50)")
-    st.markdown("- **Stage 2**: Cross-Encoder Re-ranking (Top 20)")
-    st.markdown("- **Stage 3**: BM25 Keyword Matching")
-    st.markdown("- **Stage 4**: LLM Intent Analysis")
-    st.markdown("- **Final**: Combined Scoring (Top 5)")
-    st.markdown("### 📊 Models Used")
-    st.markdown("- **Embedding**: BAAI/bge-large-en-v1.5")
-    st.markdown("- **Cross-Encoder**: ms-marco-MiniLM-L6-v2")
-    st.markdown("- **Intent Analysis**: Qwen/Qwen3-1.7B")
+    st.markdown("### 🤖 Pipeline Overview")
+    st.markdown("**5-Stage Advanced Pipeline:**")
+    st.markdown("1. FAISS Recall (Top 50)")
+    st.markdown("2. Cross-Encoder Re-ranking (Top 20)")
+    st.markdown("3. BM25 Keyword Matching")
+    st.markdown("4. LLM Intent Analysis")
+    st.markdown("5. Combined Scoring")
+    
     st.markdown("### 📈 Scoring Formula")
-    st.markdown("**Final Score = Cross-Encoder (0-1) + BM25 (0.1-0.2) + Intent (0-0.3)**")
+    st.markdown("**Final Score = Cross-Encoder (0-0.7) + BM25 (0.1-0.2) + Intent (0-0.1)**")
 
 # Initialize session state
 if 'embedding_model' not in st.session_state:
@@ -134,7 +126,7 @@ def load_cross_encoder():
     try:
         with st.spinner("🔄 Loading Cross-Encoder ms-marco-MiniLM-L6-v2..."):
             from sentence_transformers import CrossEncoder
-            # Try with explicit device specification
+            # Try with explicit device specification and logistic scoring
             device = 'cuda' if torch.cuda.is_available() else 'cpu'
             model = CrossEncoder('cross-encoder/ms-marco-MiniLM-L6-v2', device=device)
             st.success("✅ Cross-Encoder model loaded successfully!")
@@ -142,7 +134,7 @@ def load_cross_encoder():
     except Exception as e:
         st.error(f"❌ Error loading Cross-Encoder model: {str(e)}")
         try:
-            # Fallback: try without device specification
+            # Fallback: try without device specification but with logistic scoring
             st.warning("🔄 Trying Cross-Encoder without device specification...")
             model = CrossEncoder('cross-encoder/ms-marco-MiniLM-L6-v2')
             st.success("✅ Cross-Encoder model loaded (fallback)!")
@@ -285,6 +277,8 @@ class ResumeScreener:
             # Get job embedding
             job_embedding = self.get_embedding(job_description)
             
+            st.write(f"🔄 Generating embeddings for {len(resume_texts)} resumes...")
+            
             # Get resume embeddings
             resume_embeddings = []
             progress_bar = st.progress(0)
@@ -301,6 +295,8 @@ class ResumeScreener:
             
             progress_bar.empty()
             
+            st.write("🔍 Building FAISS index and searching...")
+            
             # Create FAISS index
             resume_embeddings = np.array(resume_embeddings).astype('float32')
             dimension = resume_embeddings.shape[1]
@@ -311,6 +307,9 @@ class ResumeScreener:
             job_embedding = job_embedding.reshape(1, -1).astype('float32')
             scores, indices = index.search(job_embedding, min(top_k, len(resume_texts)))
             
+            # Show completion message
+            st.write(f"✅ FAISS recall completed! Found top {min(top_k, len(resume_texts))} candidates.")
+            
             return indices[0].tolist()
             
         except Exception as e:
@@ -325,6 +324,8 @@ class ResumeScreener:
                 st.error("Cross-encoder not loaded!")
                 return [(idx, 0.0) for idx in top_50_indices[:top_k]]
             
+            st.write(f"🔄 Processing {len(top_50_indices)} candidates with Cross-Encoder...")
+            
             # Prepare pairs for cross-encoder
             pairs = []
             valid_indices = []
@@ -338,18 +339,33 @@ class ResumeScreener:
                     valid_indices.append(idx)
             
             if not pairs:
+                st.warning("No valid pairs found for cross-encoder!")
                 return [(idx, 0.0) for idx in top_50_indices[:top_k]]
             
+            st.write(f"📊 Cross-Encoder analyzing {len(pairs)} resume-job pairs...")
+            
             # Get cross-encoder scores
             progress_bar = st.progress(0)
             scores = []
             
+            def safe_sigmoid(x):
+                """Safe sigmoid function that handles overflow"""
+                if x >= 0:
+                    exp_neg_x = np.exp(-x)
+                    return 1 / (1 + exp_neg_x)
+                else:
+                    exp_x = np.exp(x)
+                    return exp_x / (1 + exp_x)
+            
             # Process in batches to avoid memory issues
             batch_size = 8
             for i in range(0, len(pairs), batch_size):
                 batch = pairs[i:i+batch_size]
+                # Get raw logits from cross-encoder
                 batch_scores = self.cross_encoder.predict(batch)
-                scores.extend(batch_scores)
+                # Apply sigmoid to convert logits to [0,1] range
+                batch_scores_sigmoid = [safe_sigmoid(score) for score in batch_scores]
+                scores.extend(batch_scores_sigmoid)
                 progress_bar.progress(min(1.0, (i + batch_size) / len(pairs)))
             
             progress_bar.empty()
@@ -358,6 +374,26 @@ class ResumeScreener:
             indexed_scores = list(zip(valid_indices, scores))
             indexed_scores.sort(key=lambda x: x[1], reverse=True)
             
+            # Normalize scores to 0-0.7 range (highest score becomes 0.7)
+            if scores and len(scores) > 0:
+                max_score = max(scores)
+                min_score = min(scores)
+                
+                if max_score > min_score:
+                    # Scale to 0-0.7 range
+                    normalized_indexed_scores = []
+                    for idx, score in indexed_scores:
+                        normalized_score = 0.7 * (score - min_score) / (max_score - min_score)
+                        normalized_indexed_scores.append((idx, normalized_score))
+                    indexed_scores = normalized_indexed_scores
+                else:
+                    # All scores are the same, give them all 0.35 (middle value)
+                    indexed_scores = [(idx, 0.35) for idx, _ in indexed_scores]
+            
+            # Show completion message
+            st.write(f"✅ Cross-Encoder completed! Selected top {min(top_k, len(indexed_scores))} candidates.")
+            st.write(f"📊 Cross-Encoder scores normalized to 0-0.7 range (highest: {indexed_scores[0][1]:.3f})")
+            
             return indexed_scores[:top_k]
             
         except Exception as e:
@@ -367,6 +403,8 @@ class ResumeScreener:
     def add_bm25_scores(self, resume_texts, job_description, top_20_results):
         """Stage 3: Add BM25 scores to top 20 resumes"""
         try:
+            st.write(f"🔄 Calculating BM25 keyword scores for {len(top_20_results)} candidates...")
+            
             # Get texts for top 20
             top_20_texts = [resume_texts[idx] for idx, _ in top_20_results]
             
@@ -393,6 +431,8 @@ class ResumeScreener:
                 bm25_score = normalized_bm25[i] if i < len(normalized_bm25) else 0.15
                 results_with_bm25.append((idx, cross_score, bm25_score))
             
+            st.write(f"✅ BM25 keyword matching completed!")
+            
             return results_with_bm25
             
         except Exception as e:
@@ -406,7 +446,12 @@ class ResumeScreener:
             progress_bar = st.progress(0)
             
             for i, (idx, cross_score, bm25_score) in enumerate(top_20_with_bm25):
-                intent_score = self.analyze_intent(resume_texts[idx], job_description)
+                candidate_name = st.session_state.file_names[idx] if idx < len(st.session_state.file_names) else f"Resume_{idx}"
+                intent_score, intent_text = self.analyze_intent(resume_texts[idx], job_description)
+                
+                # Print the intent analysis result
+                st.write(f"📋 **{candidate_name}**: Intent = **{intent_text}** (Score: {intent_score:.1f})")
+                
                 results_with_intent.append((idx, cross_score, bm25_score, intent_score))
                 progress_bar.progress((i + 1) / len(top_20_with_bm25))
             
@@ -424,49 +469,65 @@ class ResumeScreener:
             resume_snippet = resume_text[:1500] if len(resume_text) > 1500 else resume_text
             job_snippet = job_description[:800] if len(job_description) > 800 else job_description
             
-            prompt = f"""You are given a job description and a candidate's resume.
-Clearly answer: "Is the candidate likely seeking this job? Respond with 'Yes', 'Maybe', or 'No' and give a brief justification."
+            prompt = f"""You are a helpful HR assistant. Look at this candidate's resume and job posting.
 
-Job Description:
+The candidate is likely a good fit if they have ANY of these:
+- Related work experience (even if different industry)
+- Relevant technical skills
+- Educational background that could apply
+- Any transferable skills
+- Similar job titles or responsibilities
+
+Be generous in your assessment. Most candidates who made it this far are potentially suitable.
+
+Answer "Yes" for most candidates unless they are completely unrelated.
+Answer "No" only if absolutely no connection exists.
+
+Job Posting:
 {job_snippet}
 
 Candidate Resume:
 {resume_snippet}
 
-Response format:
-Intent: [Yes/Maybe/No]
-Reason: [Brief justification]"""
+Is this candidate suitable? Answer:"""
             
             response = generate_qwen3_response(
                 prompt,
                 st.session_state.qwen3_intent_tokenizer,
                 st.session_state.qwen3_intent_model,
-                max_new_tokens=100
+                max_new_tokens=20
             )
             
-            # Parse response
-            response_lower = response.lower()
-            if 'intent: yes' in response_lower or 'intent:yes' in response_lower:
-                return 0.3
-            elif 'intent: maybe' in response_lower or 'intent:maybe' in response_lower:
-                return 0.1
+            # Debug: print the raw response
+            print(f"Raw LLM response: '{response}'")
+            
+            # Parse response - look for the answer directly
+            response_lower = response.lower().strip()
+            if 'yes' in response_lower:
+                return 0.1, "Yes"
+            elif 'no' in response_lower:
+                return 0.0, "No"
             else:
-                return 0.0
+                # If no clear answer, default to "Yes" to be more lenient
+                print(f"Unclear response, defaulting to Yes: '{response}'")
+                return 0.1, "Yes"
                 
         except Exception as e:
             st.warning(f"Error analyzing intent: {str(e)}")
-            return 0.1  # Default to "Maybe"
+            return 0.1, "Yes"  # Default to "Yes" instead of "Maybe"
     
     def calculate_final_scores(self, results_with_all_scores):
         """Stage 5: Calculate final combined scores"""
         try:
+            st.write(f"🔄 Computing final combined scores for {len(results_with_all_scores)} candidates...")
+            
             final_results = []
             
             for idx, cross_score, bm25_score, intent_score in results_with_all_scores:
-                # Normalize cross-encoder score to 0-1 range
-                normalized_cross = max(0, min(1, cross_score))
+                # Cross-encoder scores are already in [0,1] range with logistic scoring
+                normalized_cross = cross_score
                 
-                # Final Score = Cross-Encoder (0-1) + BM25 (0.1-0.2) + Intent (0-0.3)
+                # Final Score = Cross-Encoder (0-0.7) + BM25 (0.1-0.2) + Intent (0-0.1)
                 final_score = normalized_cross + bm25_score + intent_score
                 
                 final_results.append({
@@ -480,52 +541,15 @@ Reason: [Brief justification]"""
             # Sort by final score
             final_results.sort(key=lambda x: x['final_score'], reverse=True)
             
+            st.write(f"✅ Final ranking completed! Candidates sorted by combined score.")
+            
             return final_results
             
         except Exception as e:
             st.error(f"Error calculating final scores: {str(e)}")
             return []
     
-    def extract_skills(self, text, job_description):
-        """Extract skills from resume based on job description"""
-        if not text:
-            return []
-            
-        # Common tech skills
-        common_skills = [
-            "python", "java", "javascript", "react", "angular", "vue", "node.js", 
-            "express", "django", "flask", "spring", "sql", "nosql", "html", "css",
-            "aws", "azure", "gcp", "docker", "kubernetes", "jenkins", "git", "github",
-            "agile", "scrum", "jira", "ci/cd", "devops", "microservices", "rest", "api",
-            "machine learning", "deep learning", "data science", "artificial intelligence", 
-            "tensorflow", "pytorch", "keras", "scikit-learn", "pandas", "numpy", 
-            "matplotlib", "seaborn", "jupyter", "r", "sas", "spss", "tableau", "powerbi",
-            "excel", "mysql", "postgresql", "mongodb", "redis", "elasticsearch", 
-            "kafka", "rabbitmq", "spark", "hadoop", "hive", "airflow", "linux", "unix"
-        ]
-        
-        # Extract potential skills from job description
-        job_words = set(word.lower() for word in word_tokenize(job_description) if len(word) > 2)
-        
-        # Find matching skills
-        found_skills = []
-        text_lower = text.lower()
-        
-        # Check common skills that appear in both resume and job description
-        for skill in common_skills:
-            if skill in text_lower and any(skill in job_word for job_word in job_words):
-                found_skills.append(skill)
-        
-        # Check for skills mentioned in job description
-        for word in job_words:
-            if len(word) > 3 and word in text_lower and word not in found_skills:
-                # Basic filter to avoid common words
-                if word not in ['with', 'have', 'that', 'this', 'from', 'what', 'when', 'where']:
-                    found_skills.append(word)
-        
-        return list(set(found_skills))[:15]  # Return top 15 unique skills
-    
-    def generate_simple_explanation(self, score, semantic_score, bm25_score, skills):
+    def generate_simple_explanation(self, score, semantic_score, bm25_score):
         """Generate simple explanation for the match (fallback)"""
         if score > 0.8:
             quality = "excellent"
@@ -543,9 +567,6 @@ Reason: [Brief justification]"""
         else:
             explanation += f"The resume has high keyword match ({bm25_score:.2f}) indicating direct skill alignment. "
         
-        if skills:
-            explanation += f"Key matching competencies include: {', '.join(skills[:5])}."
-        
         return explanation
     
 
@@ -772,9 +793,6 @@ if st.button("🚀 Advanced Pipeline Analysis",
                     name = st.session_state.file_names[idx]
                     text = st.session_state.resume_texts[idx]
                     
-                    # Extract skills
-                    skills = screener.extract_skills(text, job_description)
-                    
                     results.append({
                         'rank': rank,
                         'name': name,
@@ -782,7 +800,7 @@ if st.button("🚀 Advanced Pipeline Analysis",
                         'cross_encoder_score': result_data['cross_encoder_score'],
                         'bm25_score': result_data['bm25_score'],
                         'intent_score': result_data['intent_score'],
-                        'skills': skills,
+                        'skills': [],
                         'text': text,
                         'text_preview': text[:500] + "..." if len(text) > 500 else text,
                         'explanation': None  # Will be filled with simple explanation
@@ -793,8 +811,7 @@ if st.button("🚀 Advanced Pipeline Analysis",
                     result['explanation'] = screener.generate_simple_explanation(
                         result['final_score'], 
                         result['cross_encoder_score'], 
-                        result['bm25_score'], 
-                        result['skills']
+                        result['bm25_score']
                     )
                 
                 # Store in session state
@@ -818,7 +835,7 @@ if st.session_state.results:
         summary_data = []
         for result in st.session_state.results:
             # Map intent score to text
-            intent_text = "Yes" if result['intent_score'] == 0.3 else "Maybe" if result['intent_score'] == 0.1 else "No"
+            intent_text = "Yes" if result['intent_score'] == 0.1 else "No"
             
             summary_data.append({
                 "Rank": result['rank'],
@@ -826,8 +843,7 @@ if st.session_state.results:
                 "Final Score": f"{result['final_score']:.2f}",
                 "Cross-Encoder": f"{result['cross_encoder_score']:.2f}",
                 "BM25": f"{result['bm25_score']:.2f}",
-                "Intent": f"{intent_text} ({result['intent_score']:.1f})",
-                "Top Skills": ", ".join(result['skills'][:5])
+                "Intent": f"{intent_text} ({result['intent_score']:.1f})"
             })
         
         summary_df = pd.DataFrame(summary_data)
@@ -862,7 +878,7 @@ if st.session_state.results:
         # Download link
         detailed_data = []
         for result in st.session_state.results:
-            intent_text = "Yes" if result['intent_score'] == 0.3 else "Maybe" if result['intent_score'] == 0.1 else "No"
+            intent_text = "Yes" if result['intent_score'] == 0.1 else "No"
             
             detailed_data.append({
                 "Rank": result['rank'],
@@ -872,7 +888,6 @@ if st.session_state.results:
                 "BM25_Score": result['bm25_score'],
                 "Intent_Score": result['intent_score'],
                 "Intent_Analysis": intent_text,
-                "Skills": "; ".join(result['skills']),
                 "AI_Explanation": result['explanation'],
                 "Resume_Preview": result['text_preview']
             })
@@ -883,7 +898,7 @@ if st.session_state.results:
     with tab2:
         # Detailed results with new scoring breakdown
         for result in st.session_state.results:
-            intent_text = "Yes" if result['intent_score'] == 0.3 else "Maybe" if result['intent_score'] == 0.1 else "No"
+            intent_text = "Yes" if result['intent_score'] == 0.1 else "No"
             
             with st.expander(f"#{result['rank']}: {result['name']} (Final Score: {result['final_score']:.2f})"):
                 col1, col2 = st.columns([1, 2])
@@ -892,16 +907,9 @@ if st.session_state.results:
                     st.metric("🏆 Final Score", f"{result['final_score']:.2f}")
                     
                     st.write("**📊 Score Breakdown:**")
-                    st.metric("🎯 Cross-Encoder", f"{result['cross_encoder_score']:.2f}", help="Semantic relevance (0-1)")
+                    st.metric("🎯 Cross-Encoder", f"{result['cross_encoder_score']:.2f}", help="Semantic relevance (0-0.7)")
                     st.metric("🔤 BM25 Keywords", f"{result['bm25_score']:.2f}", help="Keyword matching (0.1-0.2)")
-                    st.metric("🤖 Intent Analysis", f"{intent_text} ({result['intent_score']:.1f})", help="Job seeking likelihood (0-0.3)")
-                    
-                    st.write("**🎯 Matching Skills:**")
-                    skills_per_column = 5
-                    skill_cols = st.columns(2)
-                    for idx, skill in enumerate(result['skills'][:10]):
-                        with skill_cols[idx % 2]:
-                            st.write(f"• {skill}")
+                    st.metric("🤖 Intent Analysis", f"{intent_text} ({result['intent_score']:.2f})", help="Job seeking likelihood (0-0.1)")
                 
                 with col2:
                     st.write("**💡 AI-Generated Match Analysis:**")