app.py

import os
import gradio as gr
import whisper
import re
import datetime
import pandas as pd
import google.generativeai as genai

# ✅ Configure Gemini API
genai.configure(api_key="AIzaSyAeNCjKJVCT0gmQRAPq4NltXkc-1zELH28")
model = genai.GenerativeModel("gemini-1.5-flash-latest")
asr_model = whisper.load_model("base")

# ✅ Global session state
session = {
    "username": "",
    "company": "",
    "role": "",
    "questions": [],
    "index": 0,
    "feedback": [],
    "interview_done": False
}

# ✅ Dummy public data with full analysis
public_data = pd.DataFrame([
    {"Username": "Flipkart employee", "Company": "amazon", "Role": "services", "Date": "2025-05-27", "Tone (%)": 60, "Vocabulary (%)": 50, "Grammar (%)": 40, "Technical (%)": 30},
    {"Username": "NIELT", "Company": "Dmart", "Role": "Marketing and Sales Funnel", "Date": "2025-05-22", "Tone (%)": 74, "Vocabulary (%)": 70, "Grammar (%)": 76, "Technical (%)": 69},
    {"Username": "AIML Saddi", "Company": "Nykaa", "Role": "Product Manager", "Date": "2025-05-27", "Tone (%)": 86, "Vocabulary (%)": 79, "Grammar (%)": 81, "Technical (%)": 88},
    {"Username": "AIML Deepanshu", "Company": "Unilever", "Role": "Brand Manager", "Date": "2025-05-27", "Tone (%)": 80, "Vocabulary (%)": 78, "Grammar (%)": 83, "Technical (%)": 82},
    {"Username": "Rahul Ropar", "Company": "Godrej", "Role": "B2B Sales", "Date": "2025-05-23", "Tone (%)": 72, "Vocabulary (%)": 75, "Grammar (%)": 69, "Technical (%)": 65},
    {"Username": "Vidit789", "Company": "Amazon", "Role": "Cloud Security Analyst", "Date": "2025-05-23", "Tone (%)": 90, "Vocabulary (%)": 82, "Grammar (%)": 87, "Technical (%)": 91},
    {"Username": "DSP Dev", "Company": "Tesla", "Role": "Autopilot QA", "Date": "2025-05-22", "Tone (%)": 84, "Vocabulary (%)": 77, "Grammar (%)": 81, "Technical (%)": 79}
])

# ✅ Start interview
def start_interview(username, company, role):
    if not username or not company or not role:
        return "❌ Please fill all fields first."
    session.update({
        "username": username,
        "company": company,
        "role": role,
        "questions": [],
        "index": 0,
        "feedback": [],
        "interview_done": False
    })
    greeting = (
        "Hello, I am Mrinankush Dutta.\n\n"
        "Welcome to your AI Mock interview session.\n\n"
        "This Mock interview provides tailored support for interviewees in real time.\n\n"
        "Just focus on being yourself. We will handle the rest."
    )
    return greeting

# ✅ Generate interview questions
def generate_questions():
    if not session["company"] or not session["role"]:
        return "❌ Please complete setup first."
    prompt = f"Generate 13 internship interview questions (8 technical and 5 behavioral) for the role of {session['role']} at {session['company']}. Number them 1 to 13."
    response = model.generate_content(prompt)
    session["questions"] = re.findall(r"^\d+\.\s+.*", response.text.strip(), re.MULTILINE)
    return next_question()

# ✅ Show next question
def next_question():
    if session["index"] >= len(session["questions"]):
        session["interview_done"] = True
        return "Interview complete. Click 'Finish' to view feedback."
    return session["questions"][session["index"]]

# ✅ Record and process answer
def process_answer(audio):
    if audio is None:
        return "❌ Please record your answer."
    result = asr_model.transcribe(audio)
    transcript = result["text"]
    q = session["questions"][session["index"]]
    session["index"] += 1
    prompt = (
        f"Interview Question: {q}\n"
        f"Answer: {transcript}\n\n"
        "Evaluate the following in percentage:\n"
        "Tone, Grammar, Vocabulary, and Technical correctness.\n"
        "Respond in this format:\n"
        "Tone: %\nGrammar: %\nVocabulary: %\nTechnical: %"
    )
    feedback = model.generate_content(prompt).text.strip()
    session["feedback"].append({"question": q, "answer": transcript, "feedback": feedback})
    return "✅ Answer recorded."

# ✅ Compile full feedback summary
def feedback_summary():
    session["interview_done"] = True
    summary = ""
    for i, entry in enumerate(session["feedback"], 1):
        summary += f"Q{i}: {entry['question']}\nAnswer: {entry['answer']}\n{entry['feedback']}\n{'-'*40}\n"
    return summary

# ✅ Public sharing after confirmation
def make_public(share_decision):
    if not session["interview_done"]:
        return "❌ Complete interview first."
    if share_decision.lower() != "yes":
        return "✅ Interview ended. Your data was not shared publicly."
    last_feedback = session["feedback"][-1]["feedback"]
    tone = re.search(r"Tone:\s*(\d+)%", last_feedback)
    grammar = re.search(r"Grammar:\s*(\d+)%", last_feedback)
    vocab = re.search(r"Vocabulary:\s*(\d+)%", last_feedback)
    tech = re.search(r"Technical:\s*(\d+)%", last_feedback)
    today = datetime.date.today().strftime("%Y-%m-%d")
    new_row = {
        "Username": session["username"],
        "Company": session["company"],
        "Role": session["role"],
        "Date": today,
        "Tone (%)": int(tone.group(1)) if tone else 70,
        "Vocabulary (%)": int(vocab.group(1)) if vocab else 70,
        "Grammar (%)": int(grammar.group(1)) if grammar else 70,
        "Technical (%)": int(tech.group(1)) if tech else 70
    }
    global public_data
    public_data.loc[len(public_data)] = new_row
    return "✅ Your analysis has been shared publicly."

# ✅ View public profiles as DataFrame
def show_public():
    if not session["interview_done"]:
        return public_data.iloc[0:0]  # Empty table if not finished
    return public_data

# ✅ Gradio UI
with gr.Blocks(theme=gr.themes.Soft()) as demo:
    gr.Markdown("# 🧠 AI MOCK INTERVIEW BUDDY")

    uname = gr.Textbox(label="👤 Unique Username")
    comp = gr.Textbox(label="🏢 Company")
    role = gr.Textbox(label="🎯 Role")
    start_btn = gr.Button("🚀 Start Interview")
    greet = gr.Textbox(label="Greeting")
    start_btn.click(start_interview, inputs=[uname, comp, role], outputs=greet)

    gen_btn = gr.Button("📋 Generate Questions")
    qbox = gr.Textbox(label="Current Question")
    gen_btn.click(generate_questions, outputs=qbox)

    record = gr.Audio(sources=["microphone"], type="filepath", label="🎤 Record Answer")
    submit = gr.Button("✅ Submit")
    result = gr.Textbox(label="Status")
    submit.click(process_answer, inputs=record, outputs=result)

    next_btn = gr.Button("➡ Next Question")
    next_btn.click(next_question, outputs=qbox)

    final = gr.Button("📊 End Interview")
    fb = gr.Textbox(label="📄 Final Feedback", lines=15)
    final.click(feedback_summary, outputs=fb)

    confirm = gr.Textbox(label="Do you want to share publicly? (Yes/No)")
    share_btn = gr.Button("☑ Share Final Results")
    status = gr.Textbox(label="Sharing Status")
    share_btn.click(make_public, inputs=confirm, outputs=status)

    view = gr.Button("📣 View Public Profiles")
    public_table = gr.Dataframe(label="🧠 Public Summary Table", interactive=False)
    view.click(show_public, outputs=public_table)

# ✅ Fixed launch block
if __name__ == "__main__":
    demo.launch()