main.py

import os
import json
import asyncio
import requests
import librosa
import soundfile as sf
from datetime import datetime
from fastapi import FastAPI, Request, Response, BackgroundTasks
from fastapi.staticfiles import StaticFiles
from dotenv import load_dotenv
from supabase import create_client, Client
from huggingface_hub import InferenceClient
from transformers import pipeline
from twilio.rest import Client as TwilioClient
from fpdf import FPDF
import edge_tts
import re

load_dotenv()

app = FastAPI()

# --- 1. CONFIGURATION ---
HF_API_KEY = os.getenv("HF_API_KEY")
SUPABASE_URL = os.getenv("SUPABASE_URL")
SUPABASE_KEY = os.getenv("SUPABASE_KEY")
TW_SID = os.getenv("TWILIO_ACCOUNT_SID")
TW_TOKEN = os.getenv("TWILIO_AUTH_TOKEN")
TW_PHONE = os.getenv("TWILIO_PHONE_NUMBER")
BASE_URL = os.getenv("BASE_URL", "").rstrip("/") 
OWNER_PHONE = os.getenv("OWNER_PHONE")

if not os.path.exists("static"): os.makedirs("static")
app.mount("/static", StaticFiles(directory="static"), name="static")

try:
    supabase: Client = create_client(SUPABASE_URL, SUPABASE_KEY)
    hf_client = InferenceClient(token=HF_API_KEY)
    twilio_client = TwilioClient(TW_SID, TW_TOKEN)
except Exception as e: print(f"⚠️ Init Error: {e}")

# --- 2. LOCAL EARS ---
print("⏳ Loading Local ASR Model...")
try:
    asr_pipeline = pipeline("automatic-speech-recognition", model="NCAIR1/NigerianAccentedEnglish", token=HF_API_KEY)
    print("✅ ASR Ready!")
except: asr_pipeline = None

BRAIN_MODEL_ID = "meta-llama/Meta-Llama-3-8B-Instruct"

MARKET_PRICES = {"cement": 5500, "rice": 80000, "yam": 2500, "pepper": 15000}

# --- 3. HELPER FUNCTIONS ---
def convert_audio_to_16k(input_path):
    try:
        y, sr = librosa.load(input_path, sr=16000)
        sf.write("temp_16k.wav", y, 16000)
        return "temp_16k.wav"
    except: return None

def transcribe_locally(file_path):
    if not asr_pipeline: return None
    try: return asr_pipeline(file_path).get("text", "")
    except: return None

def update_credit_score(user_phone):
    try:
        res = supabase.table("transactions").select("id", count="exact").eq("user_phone", user_phone).execute()
        count = res.count or 0
        new_score = min(800, 300 + (count * 10))
        supabase.table("users").upsert({"phone": user_phone, "credit_score": new_score}).execute()
        return new_score
    except: return 300

def simulate_receipt_scan():
    return "I bought 30 cartons of Indomie for 150,000 naira."

def generate_pdf_report(user_phone):
    try:
        pdf = FPDF()
        pdf.add_page()
        pdf.set_font("Arial", size=12)
        pdf.cell(200, 10, txt="AkoweAje Financial Report", ln=1, align='C')
        pdf.cell(200, 10, txt=f"Generated for: {user_phone}", ln=1, align='C')
        pdf.cell(200, 10, txt=f"Date: {datetime.now().strftime('%Y-%m-%d')}", ln=1, align='C')
        pdf.ln(10)

        score = update_credit_score(user_phone)
        pdf.set_font("Arial", 'B', 14)
        pdf.cell(200, 10, txt=f"Akowe Credit Score: {score}/800", ln=1)
        pdf.ln(5)

        res = supabase.table("transactions").select("*").order("created_at", desc=True).limit(20).execute()
        pdf.set_font("Arial", 'B', 10)
        pdf.cell(30, 10, "Date", 1)
        pdf.cell(25, 10, "Type", 1)
        pdf.cell(75, 10, "Item", 1)
        pdf.cell(35, 10, "Amount", 1)
        pdf.cell(25, 10, "Profit", 1)
        pdf.ln()
        
        pdf.set_font("Arial", size=9)
        total_sales = 0
        for row in res.data:
            date_str = row.get('created_at', '')[:10]
            intent = row.get('intent') or 'UNK'
            item = row.get('item') or 'Unknown'
            amount = row.get('amount') or 0
            profit = row.get('profit') or 0
            
            pdf.cell(30, 10, date_str, 1)
            pdf.cell(25, 10, intent[:10], 1)
            pdf.cell(75, 10, item[:35], 1)
            pdf.cell(35, 10, f"N{amount:,}", 1)
            pdf.cell(25, 10, f"N{profit:,}", 1)
            pdf.ln()
            if intent == 'SALE': total_sales += amount

        pdf.ln(10)
        pdf.set_font("Arial", 'B', 12)
        pdf.cell(200, 10, txt=f"Total Revenue: N{total_sales:,.2f}", ln=1)

        filename = f"static/report_{int(datetime.now().timestamp())}.pdf"
        pdf.output(filename)
        return filename
    except Exception as e:
        print(f"❌ PDF Error: {e}")
        return None

# --- VOICE BRIEFING (DUAL-ENGINE FALLBACK) ---
async def generate_voice_briefing(text):
    filename = f"static/briefing_{int(datetime.now().timestamp())}.mp3"
    
    # Engine 1: Nigerian Accent (Preferred)
    try:
        print("🎙️ Attempting Nigerian Voice...")
        communicate = edge_tts.Communicate(text, "en-NG-EzinneNeural")
        await communicate.save(filename)
        return filename
    except Exception as e:
        print(f"⚠️ Nigerian Voice Failed: {e}")

    # Engine 2: US Accent (Fallback)
    try:
        print("🎙️ Attempting US Fallback Voice...")
        communicate = edge_tts.Communicate(text, "en-US-AriaNeural")
        await communicate.save(filename)
        return filename
    except Exception as e:
        print(f"❌ All TTS Failed: {e}")
        return None

# --- OMNI-BRAIN ---
def extract_financial_data(text_input):
    system_prompt = """You are AkoweAje, a strict data extraction AI. 
    Task: Convert the user's input into valid JSON.
    Do NOT speak. Do NOT apologize. OUTPUT JSON ONLY.
    
    INTENTS:
    - 'SALE': New sale (extract item, amount, cost_price).
    - 'DEBT': New debt (extract customer, amount).
    - 'GENERATE_REPORT': User wants a PDF.
    - 'GENERATE_REMINDER': User wants a text message for a debtor.
    - 'DAILY_BRIEF': User wants a voice summary.
    - 'GENERATE_AD': User wants marketing text.
    - 'QUERY_SCORE': User asks for credit score.

    EXAMPLES:
    Input: "I sell 50 bags of cement to Dangote for 200k. I buy am 150k."
    JSON: {"intent": "SALE", "item": "50 bags of cement", "amount": 200000, "cost_price": 150000, "customer": "Dangote"}
    
    Input: "Write a message to dangote to pay me."
    JSON: {"intent": "GENERATE_REMINDER", "customer": "Dangote"}

    Input: "Write advert for fresh pepper"
    JSON: {"intent": "GENERATE_AD", "item": "fresh pepper"}

    Input: "{text_input}"
    JSON:"""
    
    messages = [{"role": "system", "content": system_prompt}, {"role": "user", "content": text_input}]
    try:
        response = hf_client.chat_completion(model=BRAIN_MODEL_ID, messages=messages, max_tokens=200, temperature=0.1)
        raw_content = response.choices[0].message.content.strip()
        
        start = raw_content.find("{")
        end = raw_content.rfind("}")
        if start != -1 and end != -1:
            json_str = raw_content[start:end+1]
            return json.loads(json_str)
    except Exception as e:
        print(f"❌ Brain Error: {e}")
    return {"intent": "UNKNOWN"}

# --- MAIN PROCESSOR ---
async def process_message_task(sender_phone, body, media_url, media_type):
    final_text = body
    is_image_scan = False

    if media_url and media_type and media_type.startswith('image/'):
        print("📷 Image received. Simulating scan...")
        final_text = simulate_receipt_scan()
        is_image_scan = True

    elif media_url and media_type and media_type.startswith('audio/'):
        try:
            r = requests.get(media_url, headers={'User-Agent': 'Mozilla/5.0'})
            with open("temp_voice.ogg", "wb") as f: f.write(r.content)
            wav = convert_audio_to_16k("temp_voice.ogg")
            if wav: final_text = transcribe_locally(wav) or final_text
        except: pass

    if not final_text: return
    print(f"📝 Analyzing: {final_text}")
    
    data = extract_financial_data(final_text)
    intent = data.get("intent")
    reply_msg = ""
    
    if intent in ["SALE", "DEBT"] and OWNER_PHONE and sender_phone != OWNER_PHONE:
        alert_msg = f"🚨 SECURITY ALERT: Apprentice ({sender_phone}) just logged: {final_text}"
        try: twilio_client.messages.create(from_=TW_PHONE, to=OWNER_PHONE, body=alert_msg)
        except: pass

    if intent == "SALE":
        amount = int(data.get('amount') or 0)
        cost_price = int(data.get('cost_price') or 0)
        item = data.get('item') or "Unknown Item"
        
        profit_msg = ""
        profit = 0
        if cost_price > 0 and amount > 0:
            profit = amount - cost_price
            margin = (profit / amount) * 100
            profit_msg = f"\n📈 Profit: ₦{profit:,.0f} ({margin:.1f}%)"
        
        watchdog_msg = ""
        if amount > 0:
            for key, mkt_price in MARKET_PRICES.items():
                if key in item.lower():
                    if amount < mkt_price:
                        watchdog_msg = f"\n⚠️ Market Alert: You are selling {key} cheap! Market price is N{mkt_price}."

        try:
            supabase.table("transactions").insert({
                "user_phone": sender_phone, "raw_text": final_text, "intent": "SALE",
                "item": item, "amount": amount,
                "cost_price": cost_price, "profit": profit
            }).execute()
        except Exception as e: print(f"DB Error: {e}")

        prefix = "📷 Receipt Scanned &" if is_image_scan else "💰"
        reply_msg = f"{prefix} Sale Recorded!\nItem: {item}\nAmount: ₦{amount:,.0f}{profit_msg}{watchdog_msg}"
        update_credit_score(sender_phone)

    elif intent == "GENERATE_AD":
        reply_msg = f"🚨 FRESH STOCK ALERT! 🚨\nJust landed: {data.get('item')}! Best quality, best price. Hurry before it finishes! 🏃🏾‍♂️💨"

    elif intent == "QUERY_SCORE":
        score = update_credit_score(sender_phone)
        reply_msg = f"🏆 Your current Akowe Credit Score is: {score}/800.\nKeep recording daily to increase it!"

    elif intent == "GENERATE_REMINDER":
        customer = data.get('customer', 'Customer')
        reply_msg = f"📝 Draft Message:\n\n'Hello {customer}, gentle reminder about the outstanding payment. Please let me know when to expect it. Thanks!'"

    elif intent == "GENERATE_REPORT":
        pdf_path = generate_pdf_report(sender_phone)
        if pdf_path: reply_msg = f"📄 Official Financial Report & Credit Score:\n{BASE_URL}/{pdf_path}"
        else: reply_msg = "⚠️ PDF Generator is temporarily busy. Try again later."

    elif intent == "DAILY_BRIEF":
        summary = "Good evening. You had a busy day. Your credit score is rising. Check your report for details."
        mp3_path = await generate_voice_briefing(summary)
        if mp3_path: reply_msg = f"🎙️ Daily Voice Briefing:\n{BASE_URL}/{mp3_path}"
        else: reply_msg = f"🎙️ (Voice Unavailable) Daily Briefing:\n{summary}"
        
    else:
        reply_msg = f"I heard: '{final_text}'\nBut didn't understand. Try again."

    # --- BRANDING FOOTER FIX ---
    if reply_msg:
        reply_msg += "\n\n(Powered by Awarri)"

    # SEND REPLY
    try:
        twilio_client.messages.create(from_=TW_PHONE, to=sender_phone, body=reply_msg)
        print("✅ Message Sent!")
    except Exception as e: print(f"❌ Twilio Error: {e}")

@app.post("/webhook")
async def whatsapp_webhook(request: Request, background_tasks: BackgroundTasks):
    form = await request.form()
    media_type = form.get('MediaContentType0', '')
    background_tasks.add_task(process_message_task, form.get('From'), form.get('Body'), form.get('MediaUrl0'), media_type)
    return Response(content="<Response></Response>", media_type="application/xml")