Update app.py

app.py

@@ -1,3 +1,6 @@
+
+import os
+hf_token = os.getenv("Gem") # Store API token in .env
 import streamlit as st
 import os
 import logging
@@ -8,7 +11,10 @@ from langchain_community.vectorstores import FAISS
 from langchain_community.embeddings import HuggingFaceEmbeddings
 from langchain.prompts import PromptTemplate
 from langchain.llms import HuggingFaceHub
-import os
+import random
+from sklearn.feature_extraction.text import TfidfVectorizer
+from sklearn.metrics.pairwise import cosine_similarity
+
 dotenv.load_dotenv()
 
 # Load configuration from YAML
@@ -17,7 +23,6 @@ def load_config():
         return yaml.safe_load(f)
 
 config = load_config()
-hf_token = os.getenv("Gem") # Store API token in .env
 logging.basicConfig(level=logging.INFO)
 
 # Load embedding model
@@ -31,31 +36,66 @@ def extract_text_from_pdf(file):
         text += page.extract_text() or ""
     return text.strip()
 
+# Function to calculate matching score between job description and resume
+def calculate_matching_score(jd_text, resume_text):
+    vectorizer = TfidfVectorizer().fit_transform([jd_text, resume_text])
+    score = cosine_similarity(vectorizer[0], vectorizer[1])[0][0] * 100
+    return round(score, 2)
+
+# Function to generate final score based on user responses
+def calculate_final_score(responses):
+    total_questions = len(responses)
+    correct_responses = sum(1 for response in responses if "good" in response.lower() or "correct" in response.lower())
+    return round((correct_responses / total_questions) * 100, 2) if total_questions > 0 else 0
+
 # Get interview questions and assess responses
-def get_interview_response(jd_text, resume_text, candidate_response=None):
-    prompt_template = """
-    You are an AI interviewer assessing a candidate for a job role.
-    
+def get_interview_response(jd_text, resume_text, candidate_response=None, round_stage="intro", question_count=0):
+    technical_names = ["Alex", "Jordan", "Casey", "Morgan"]
+    hr_names = ["Taylor", "Jamie", "Riley", "Sam"]
+
+    if round_stage in ["technical", "coding"]:
+        interviewer_name = random.choice(technical_names)
+        role = "Technical Lead"
+    else:
+        interviewer_name = random.choice(hr_names)
+        role = "HR Manager"
+
+    prompt_template = f"""
+    My name is {interviewer_name}, and I am your {role} for this round.
+
     JOB DESCRIPTION:
     {jd_text}
 
     CANDIDATE PROFILE:
     {resume_text}
 
-    1. Start by asking an **introductory question**: "Tell me about yourself."
-    2. Then, based on the job description, ask a **technical question**.
-    3. If the candidate has already responded, evaluate their answer and provide constructive feedback.
+    This is question {question_count+1} of 5.
 
-    Maintain a professional yet friendly tone.
     """
 
-    prompt = PromptTemplate(
-        input_variables=["jd_text", "resume_text"], 
-        template=prompt_template
-    ).format(jd_text=jd_text, resume_text=resume_text)
+    if question_count >= 5:
+        return f"{interviewer_name}: This round is complete. Moving to the next stage."
+
+    if round_stage == "intro":
+        prompt_template += f"{interviewer_name}: Let's start with an introduction. Tell me about yourself."
+
+    elif round_stage == "technical":
+        prompt_template += f"{interviewer_name}: Based on your resume and the job description, here is a technical question for you."
+
+    elif round_stage == "coding":
+        prompt_template += f"{interviewer_name}: Let's move to a coding problem relevant to your role."
+
+    elif round_stage == "hr":
+        prompt_template += f"{interviewer_name}: Now let's discuss some HR aspects, starting with your motivation for this role."
+
+    elif round_stage == "final_feedback":
+        prompt_template += "Summarize the candidate’s performance in both rounds in a structured format."
 
     if candidate_response:
-        prompt += f"\n\nCANDIDATE RESPONSE: {candidate_response}\n\nAssess the response and provide feedback."
+        if candidate_response.lower() == "hint":
+            prompt_template += f"{interviewer_name}: Here is a helpful hint."
+        else:
+            prompt_template += f"The candidate answered: {candidate_response}. Assess the response and move to the next question."
 
     llm = HuggingFaceHub(
         repo_id=config["model_name"],
@@ -63,7 +103,17 @@ def get_interview_response(jd_text, resume_text, candidate_response=None):
         huggingfacehub_api_token=hf_token
     )
 
-    return llm(prompt).strip()
+    response = llm(prompt_template).strip()
+
+    # Store the full assessment in a text file for admin review
+    with open("candidate_assessment.txt", "a") as f:
+        f.write(f"Round: {round_stage}, Question {question_count+1}\n")
+        f.write(f"Interviewer: {interviewer_name} ({role})\n")
+        f.write(f"Question: {prompt_template}\n")
+        f.write(f"Candidate Response: {candidate_response}\n")
+        f.write(f"Feedback: {response}\n\n")
+
+    return response if round_stage != "final_feedback" else f"{interviewer_name}: The interview is now complete."
 
 # Streamlit UI
 st.set_page_config(page_title="AI Interviewer", layout="centered")
@@ -78,18 +128,43 @@ if jd_file and resume_file:
     jd_text = extract_text_from_pdf(jd_file)
     resume_text = extract_text_from_pdf(resume_file)
 
+    # Calculate matching score
+    matching_score = calculate_matching_score(jd_text, resume_text)
+
+    # Store interview history & matching score
     if "interview_history" not in st.session_state:
         st.session_state["interview_history"] = []
+        st.session_state["responses"] = []
         first_question = get_interview_response(jd_text, resume_text)
         st.session_state["interview_history"].append(("AI", first_question))
 
+    st.write(f"**Matching Score:** {matching_score}%")
+
     for role, msg in st.session_state["interview_history"]:
         st.chat_message(role).write(msg)
 
     query = st.chat_input("Your Response:")
-    
+
     if query:
         response = get_interview_response(jd_text, resume_text, query)
         st.session_state["interview_history"].append(("You", query))
         st.session_state["interview_history"].append(("AI", response))
-        st.rerun()
+        st.session_state["responses"].append(response)  # Store responses for final score
+        st.rerun()
+
+# Generate final score and store the results for download
+if "responses" in st.session_state and len(st.session_state["responses"]) >= 5:
+    final_score = calculate_final_score(st.session_state["responses"])
+
+    # Store all results in a text file
+    file_path = "candidate_assessment.txt"
+    with open(file_path, "w") as f:
+        f.write(f"Matching Score: {matching_score}%\n")
+        f.write(f"Final Score: {final_score}%\n\n")
+        f.write("Interview Assessment:\n")
+        for role, msg in st.session_state["interview_history"]:
+            f.write(f"{role}: {msg}\n")
+
+    # Provide file download option
+    with open(file_path, "rb") as f:
+        st.download_button("Download Assessment", f, file_name="candidate_assessment.txt")