app.py

import pandas as pd
import gradio as gr
from sentence_transformers import SentenceTransformer, util

# Load dataset
df_internships = pd.read_csv("internships.csv")

# Load lightweight embedding model
model = SentenceTransformer('all-MiniLM-L6-v2')

def recommend_internships(education, skills, sector, location):
    # 1. Candidate profile as a single string
    candidate_text = f"Education: {education}. Skills: {', '.join(skills)}. Sector: {sector}. Location: {location}"
    
    # 2. Compute embeddings
    candidate_emb = model.encode(candidate_text, convert_to_tensor=True)
    internship_texts = df_internships['Skills'] + " " + df_internships['Sector'] + " " + df_internships['Location']
    internship_embs = model.encode(list(internship_texts), convert_to_tensor=True)
    
    # 3. Compute cosine similarity
    cosine_scores = util.cos_sim(candidate_emb, internship_embs)[0]
    
    # 4. Add score to dataframe
    df_internships['Match Score'] = cosine_scores.cpu().numpy()
    
    # 5. Filter by sector/location preference (optional fallback)
    filtered = df_internships[
        (df_internships['Sector'] == sector) &
        (df_internships['Location'] == location)
    ]
    if len(filtered) < 3:
        filtered = df_internships  # fallback to all internships
    
    # 6. Sort by match score and pick top 5
    top = filtered.sort_values(by='Match Score', ascending=False).head(5)
    
    # 7. Prepare card-like recommendations
    recommendations = []
    for _, row in top.iterrows():
        card = (
            f"**{row['Title']}**\n"
            f"Location: {row['Location']}\n"
            f"Sector: {row['Sector']}\n"
            f"Skills Required: {row['Skills']}\n"
            f"Match Score: {int(row['Match Score']*100)}%"
        )
        recommendations.append(card)
    
    # 8. Save Excel for download
    top[['Title', 'Skills', 'Sector', 'Location', 'Match Score']].to_excel("output.xlsx", index=False)
    
    return recommendations, "output.xlsx"

# Gradio UI
with gr.Blocks() as demo:
    gr.Markdown("## 🌟 Skill-to-Opportunity AI Navigator for PM Internship Scheme")
    gr.Markdown("Enter your details and get top 3–5 personalized internship recommendations.")
    
    with gr.Row():
        education_input = gr.Dropdown(["High School", "Undergraduate", "Graduate", "Other"], label="Education")
        sector_input = gr.Dropdown(["Tech", "Governance", "NGOs"], label="Sector Interest")
        location_input = gr.Dropdown(["Delhi", "Bangalore", "Mumbai", "Chennai", "Hyderabad"], label="Preferred Location")
    
    skills_input = gr.CheckboxGroup(
        ["Python","Excel","SQL","Writing","Editing","Research","Communication","Creativity","Marketing","Organizing","Planning"],
        label="Select Your Skills"
    )
    
    recommend_btn = gr.Button("Get Recommendations")
    
    recommendations_output = gr.Textbox(label="Top Internship Recommendations", lines=10)
    download_output = gr.File(label="Download Excel with Top Recommendations")
    
    recommend_btn.click(
        recommend_internships,
        inputs=[education_input, skills_input, sector_input, location_input],
        outputs=[recommendations_output, download_output]
    )

if __name__ == "__main__":
    demo.launch()