app.py

import streamlit as st
import pandas as pd
import pymongo
from sentence_transformers import SentenceTransformer
import numpy as np
from sklearn.feature_extraction.text import TfidfVectorizer
import pickle
import os
from datetime import datetime

# MongoDB Connection
MONGO_URI = "mongodb://muhammadbinimran1001:[email protected]:27017,dsm-shard-00-01.inrzs.mongodb.net:27017,dsm-shard-00-02.inrzs.mongodb.net:27017/?ssl=true&replicaSet=atlas-nbg4er-shard-0&authSource=admin&retryWrites=true&w=majority"
client = pymongo.MongoClient(MONGO_URI)
db = client['test']
users_collection = db['users']
jobs_collection = db['jobs']
courses_collection = db['courses']

# Load Datasets
@st.cache_data
def load_data():
    questions_df = pd.read_csv("Generated_Skill-Based_Questions.csv")
    jobs_df = pd.read_csv("Updated_Job_Posting_Dataset.csv")
    courses_df = pd.read_csv("coursera_course_dataset_v2_no_null.csv")
    return questions_df, jobs_df, courses_df

questions_df, jobs_df, courses_df = load_data()

# Load or Initialize Model
@st.cache_resource
def load_model():
    return SentenceTransformer('all-MiniLM-L6-v2')

model = load_model()
tfidf_vectorizer = TfidfVectorizer(stop_words='english')

# Skill Extraction and Question Generation
def get_user_skills(user_id):
    user = users_collection.find_one({"_id": user_id})
    return user.get("skills", []) if user else []

def get_questions_for_skills(skills):
    questions = []
    for skill in skills:
        skill_questions = questions_df[questions_df['Skill'] == skill].sample(1)
        if not skill_questions.empty:
            questions.append(skill_questions.iloc[0])
    return pd.DataFrame(questions) if questions else None

# Answer Evaluation
def evaluate_answer(user_answer, expected_answer):
    user_embedding = model.encode([user_answer], convert_to_tensor=True)[0]
    expected_embedding = model.encode([expected_answer], convert_to_tensor=True)[0]
    score = util.pytorch_cos_sim(user_embedding, expected_embedding).item() * 100
    return max(0, round(score, 2))

# Recommendation Logic
def recommend_courses(skills, user_level):
    skill_indices = [questions_df.index[questions_df['Skill'] == skill].tolist()[0] for skill in skills if skill in questions_df['Skill'].values]
    if not skill_indices:
        return []
    course_skills = courses_df['skills'].fillna("").tolist()
    course_embeddings = model.encode(course_skills, convert_to_tensor=True)
    skill_embeddings = model.encode(skills, convert_to_tensor=True)
    similarities = util.pytorch_cos_sim(skill_embeddings, course_embeddings).cpu().numpy()
    total_scores = 0.6 * np.max(similarities, axis=0)
    idx = np.argsort(-total_scores)[:3]
    return courses_df.iloc[idx][['course_title', 'Organization']].values.tolist()

def recommend_jobs(skills, user_level):
    skill_indices = [questions_df.index[questions_df['Skill'] == skill].tolist()[0] for skill in skills if skill in questions_df['Skill'].values]
    if not skill_indices:
        return []
    job_skills = jobs_df['required_skills'].fillna("").tolist()
    job_embeddings = model.encode(job_skills, convert_to_tensor=True)
    skill_embeddings = model.encode(skills, convert_to_tensor=True)
    similarities = util.pytorch_cos_sim(skill_embeddings, job_embeddings).cpu().numpy()
    total_scores = 0.5 * np.max(similarities, axis=0)
    idx = np.argsort(-total_scores)[:3]
    return jobs_df.iloc[idx][['job_title', 'company_name', 'location']].values.tolist()

# Streamlit App
st.title("Skill Assessment & Recommendation")

# Simulate User Signup (for demo, replace with actual auth)
if 'user_id' not in st.session_state:
    st.session_state.user_id = "68233a6b7c0fd8f9d6994e"  # Example user ID
    st.session_state.skills = get_user_skills(st.session_state.user_id)
    st.session_state.scores = {}

if not st.session_state.skills:
    st.write("No skills found. Please update your profile with skills during signup.")
else:
    st.write(f"Detected Skills: {st.session_state.skills}")

    if 'questions' not in st.session_state:
        st.session_state.questions = get_questions_for_skills(st.session_state.skills)
        if st.session_state.questions is not None:
            st.session_state.questions = st.session_state.questions.reset_index(drop=True)

    if st.session_state.questions is not None and not st.session_state.questions.empty:
        for idx, row in st.session_state.questions.iterrows():
            st.subheader(f"Question for {row['Skill']}")
            user_answer = st.text_area(f"Question: {row['Question']}", key=f"answer_{idx}")
            if st.button(f"Submit Answer for {row['Skill']}", key=f"submit_{idx}"):
                score = evaluate_answer(user_answer, row['Answer'])
                st.session_state.scores[row['Skill']] = score
                st.success(f"Score for {row['Skill']}: {score}%")

        if all(skill in st.session_state.scores for skill in st.session_state.skills):
            st.write("Assessment Complete!")
            mean_score = np.mean(list(st.session_state.scores.values()))
            weak_skills = [s for s, score in st.session_state.scores.items() if score < 60]
            st.write(f"Mean Score: {mean_score:.2f}%")
            st.write(f"Weak Skills: {weak_skills}")

            courses = recommend_courses(weak_skills or st.session_state.skills, "Intermediate")
            jobs = recommend_jobs(st.session_state.skills, "Intermediate")
            st.write("Recommended Courses:", courses)
            st.write("Recommended Jobs:", jobs)

            # Update user score in MongoDB (simplified)
            users_collection.update_one(
                {"_id": st.session_state.user_id},
                {"$set": {"skills_scores": st.session_state.scores}}
            )
            st.session_state.pop('questions', None)
    else:
        st.write("No questions available for the detected skills.")

# Redirect to Dashboard (simulated)
if st.button("Go to Dashboard"):
    st.write("Redirecting to User Dashboard...")
    # In a real app, use st.experimental_rerun() or a navigation library