main.py

# uv run main.py
# http://127.0.0.1:5000/caption
import time
import os
import difflib
from flask import Flask, request, jsonify
from flask_cors import CORS
from openai import OpenAI
from dotenv import load_dotenv

load_dotenv()

app = Flask(__name__)
CORS(app)

# ==========================================
# ⚙️  CONFIGURATION ZONE (PICK ONE!)
# ==========================================

# --- OPTION 1: GOOGLE GEMINI (Fast & Free Tier) ---
# CURRENT_PROVIDER = "gemini"
# API_KEY = os.environ.get("GEMINI_API_KEY")
# BASE_URL = "https://generativelanguage.googleapis.com/v1beta/openai/"
# MODEL_NAME = "gemini-2.5-flash"

# https://dashboard.sarvam.ai/
#--- OPTION 2: SARVAM AI (Indian AI) ---
CURRENT_PROVIDER = "sarvam"
API_KEY = os.environ.get("SARVAM_API_KEY")
BASE_URL = "https://api.sarvam.ai/v1"
MODEL_NAME = "sarvam-m" 

# --- OPTION 3: PERPLEXITY (Live Search Knowledge) ---
# CURRENT_PROVIDER = "perplexity"
# API_KEY = os.environ.get("PERPLEXITY_API_KEY")
# BASE_URL = "https://api.perplexity.ai"
# MODEL_NAME = "llama-3.1-sonar-small-128k-online"

# ==========================================

# Initialize the Universal Client
client = OpenAI(
    api_key=API_KEY,
    base_url=BASE_URL
)

MIN_CHAR_LENGTH = 5  # Don't send text shorter than this (e.g. "Hello?")

# --- UI GARBAGE TO IGNORE ---
IGNORE_PHRASES = [
    "turn on microphone", "turn off microphone", 
    "turn on camera", "turn off camera",
    "share screen", "stop sharing",
    "raise hand", "lower hand",
    "caption settings", "font size",
    "meeting details", "people", "chat",
    "you are muted", "microphone is off"
]


def load_file(filename):
    try:
        with open(filename, 'r', encoding='utf-8') as f:
            return f.read().strip()
    except FileNotFoundError:
        return ""

RESUME = load_file("resume.txt")
BASE_PROMPT = load_file("system_prompt.txt")

SYSTEM_INSTRUCTION = f"""
{BASE_PROMPT}

MY RESUME:
{RESUME}

INSTRUCTIONS:
1. You are a fast, expert technical assistant.
2. Receive the raw transcript of an interviewer's question.
3. If the input is incomplete or just noise, reply with "..."
4. Otherwise, answer technically and confidently in the first person.
5. Keep it under 3 sentences.
"""

# --- GLOBAL MEMORY ---
last_processed_text = ""
last_successful_reply = ""

def is_garbage(text):
    """Returns True if text is UI noise or too short."""
    text_lower = text.lower()
    
    # 1. Length Check
    if len(text) < MIN_CHAR_LENGTH:
        print(f"[IGNORED] Too short ({len(text)} chars): {text}")
        return True

    # 2. UI Keyword Check
    if any(phrase in text_lower for phrase in IGNORE_PHRASES):
        print(f"[IGNORED] UI Garbage: {text}")
        return True
        
    return False

@app.route('/caption', methods=['POST'])
def receive_caption():
    global last_processed_text, last_successful_reply
    
    data = request.json
    incoming_text = data.get('text', '').strip()

    # --- 1. DUPLICATE CHECK ---
    # If we just processed this exact text, ignore it.
    if incoming_text == last_processed_text:
        return jsonify({"status": "ignored", "reason": "duplicate"}), 200

    last_processed_text = incoming_text

    # --- 2. GARBAGE CHECK ---
    if is_garbage(incoming_text):
        return jsonify({"status": "ignored", "reason": "garbage"}), 200

    print(f"\n[INTERVIEWER]: {incoming_text}")

    try:
        start_time = time.time()
        
        # --- 3. CALL GEMINI ---
        response = client.chat.completions.create(
            model=MODEL_NAME,
            messages=[
                {"role": "system", "content": SYSTEM_INSTRUCTION},
                {"role": "user", "content": incoming_text}
            ],
            stream=False
        )
        
        reply = response.choices[0].message.content.strip()
        latency = time.time() - start_time

        # --- 4. EMPTY RESPONSE CHECK ---
        # If AI replies with "..." or empty, don't show it
        if not reply or reply == "..." or len(reply) < 2:
            print(f"[AI DECIDED TO IGNORE] ({latency:.2f}s)")
            return jsonify({"status": "ignored", "reason": "ai_silence"}), 200

        print("="*50)
        print(f"GEMINI ({latency:.2f}s): {reply}")
        print("="*50 + "\n")
        
        last_successful_reply = reply
        return jsonify({"status": "success", "reply": reply})

    except Exception as e:
        print(f"Error: {e}")
        return jsonify({"status": "error"}), 500

if __name__ == '__main__':
    print(f"--- SECURE SERVER READY ({MODEL_NAME}) ---")
    print(f"Min Text Length: {MIN_CHAR_LENGTH} chars")
    app.run(port=5000)
    #http://127.0.0.1:5000/caption