app.py

import os
import time
import base64
import random
import json
import requests
from datetime import datetime, timedelta, timezone
from flask import Flask, request, jsonify, Response

app = Flask(__name__)
app.secret_key = os.getenv("FLASK_SECRET_KEY")

# ==== API KEYS ====
GROQ_API_KEY_1 = os.getenv("GROQ_API_KEY_1")
GROQ_API_KEY_2 = os.getenv("GROQ_API_KEY_2")
GROQ_API_KEY_3 = os.getenv("GROQ_API_KEY_3")
SERPAPI_KEY = os.getenv("SERPAPI_KEY")

# ==== URL ====
GROQ_URL_CHAT = "https://api.groq.com/openai/v1/chat/completions"
GROQ_URL_TTS = "https://api.groq.com/openai/v1/audio/speech"
GROQ_URL_STT = "https://api.groq.com/openai/v1/audio/transcriptions"

# ==== SYSTEM PROMPT ====
SYSTEM_PROMPT = (
    "You are Talk GTE — a friendly AI assistant created by Vibow AI. "
    "GTE means Generative Text Expert in Vibow AI. "
    "Vibow AI created in 29 June 2025 and Talk GTE created in 23 October 2025. "
    "Talk GTE have approximately 1 trillion parameters. "
    "Stay positive, kind, and expert. "
    "Always capitalize the first letter of sentences. "
    "If the user requests code, always use triple backticks (```). "
    "Be concise, neutral, and accurate. "
    "Sometimes use emoji but relevant."
)

# =========================
# 🎤 STT
# =========================
def transcribe_audio(file_path: str) -> str:
    try:
        headers = {"Authorization": f"Bearer {GROQ_API_KEY_2}"}
        files = {
            "file": (os.path.basename(file_path), open(file_path, "rb"), "audio/wav"),
            "model": (None, "whisper-large-v3-turbo"),
        }
        res = requests.post(GROQ_URL_STT, headers=headers, files=files, timeout=60)
        res.raise_for_status()
        return res.json().get("text", "")
    except Exception as e:
        print(f"[STT Error] {e}")
        return ""
    finally:
        if os.path.exists(file_path):
            os.remove(file_path)

# =========================
# 🔊 TTS
# =========================
def text_to_speech(text: str) -> bytes:
    try:
        headers = {"Authorization": f"Bearer {GROQ_API_KEY_3}"}
        data = {"model": "playai-tts", "voice": "Celeste-PlayAI", "input": text}
        res = requests.post(GROQ_URL_TTS, headers=headers, json=data, timeout=60)
        if res.status_code != 200:
            print(f"[TTS Error] {res.text}")
            return b""
        return res.content
    except Exception as e:
        print(f"[TTS Exception] {e}")
        return b""

# =========================
# 🌐 SerpAPI with 3 images
# =========================
def serpapi_search(query: str, location=None, num_results=3):
    url = "https://serpapi.com/search.json"
    params = {
        "q": query,
        "location": location or "",
        "engine": "google",
        "api_key": SERPAPI_KEY,
        "num": num_results
    }
    try:
        r = requests.get(url, params=params, timeout=10)
        r.raise_for_status()
        data = r.json()

        results = []
        images = []

        if "organic_results" in data:
            for item in data["organic_results"][:num_results]:
                results.append({
                    "title": item.get("title", ""),
                    "snippet": item.get("snippet", ""),
                    "link": item.get("link", "")
                })

        if "images_results" in data:
            for img in data["images_results"][:3]:
                images.append(img.get("thumbnail", ""))

        return {"results": results, "images": images}

    except Exception as e:
        print(f"[SerpAPI Error] {e}")
        return {"results": [], "images": []}

# =========================
# 💬 Stream Chat
# =========================
def stream_chat(prompt: str, history=None):
    wib = timezone(timedelta(hours=7))
    now = datetime.now(wib)
    sys_prompt = SYSTEM_PROMPT + f"\nCurrent time: {now.strftime('%A, %d %B %Y — %H:%M:%S WIB')}."

    messages = [{"role": "system", "content": sys_prompt}]
    if history:
        messages += history
    messages.append({"role": "user", "content": prompt})

    payload = {
        "model": "moonshotai/kimi-k2-instruct-0905",
        "messages": messages,
        "temperature": 0.7,
        "max_tokens": 3600,
        "stream": True,
    }

    headers = {"Authorization": f"Bearer {GROQ_API_KEY_1}"}

    with requests.post(GROQ_URL_CHAT, headers=headers, json=payload, stream=True) as r:
        for line in r.iter_lines():
            if not line:
                continue
            line = line.decode()
            if line.startswith("data: "):
                data = line[6:]
                if data == "[DONE]":
                    break
                try:
                    delta = json.loads(data)["choices"][0]["delta"].get("content", "")
                    if delta:
                        yield delta
                except:
                    continue

# =========================
# 🚀 Chat Endpoint (Text + Voice)
# =========================
@app.route("/chat", methods=["POST"])
def chat():
    # 🎤 Voice Mode
    if "audio" in request.files:
        audio = request.files["audio"]
        temp = f"/tmp/{time.time()}_{random.randint(1000,9999)}.wav"
        audio.save(temp)

        user_text = transcribe_audio(temp)
        if not user_text:
            return jsonify({"error": "Failed STT"}), 500

        keywords = ["hotel", "mall", "resort", "villa", "tempat wisata"]

        if any(k in user_text.lower() for k in keywords):
            serp = serpapi_search(user_text)
            text = "\n".join([f"{r['title']} — {r['snippet']} — {r['link']}" for r in serp["results"]])
            imgs = " ".join(serp["images"])
            user_text = f"{user_text}\n\nGoogle Results:\n{text}\n\nImages:\n{imgs}\n\nExplain & recommend."

        ai = "".join(chunk for chunk in stream_chat(user_text))
        audio_bytes = text_to_speech(ai)

        return jsonify({
            "mode": "voice",
            "transcript": user_text,
            "reply_text": ai,
            "audio_base64": "data:audio/mp3;base64," + base64.b64encode(audio_bytes).decode()
        })

    # 💬 Text Mode
    data = request.get_json(force=True)
    prompt = data.get("prompt", "")
    history = data.get("history", [])

    keywords = ["hotel", "mall", "resort", "villa", "tempat wisata"]

    if any(k in prompt.lower() for k in keywords):
        serp = serpapi_search(prompt)
        text = "\n".join([f"{r['title']} — {r['snippet']} — {r['link']}" for r in serp["results"]])
        imgs = " ".join(serp["images"])
        prompt = f"{prompt}\n\nGoogle Results:\n{text}\n\nImages:\n{imgs}\n\nExplain & recommend."

    def generate():
        for chunk in stream_chat(prompt, history):
            yield chunk

    return Response(generate(), mimetype="text/plain")

# =========================
# ▶️ Run
# =========================
if __name__ == "__main__":
    print("🚀 Vibow Talk GTE Server Running")
    app.run(host="0.0.0.0", port=7860, debug=True, threaded=True)