ai.py

import pytz
import requests
import time
from datetime import datetime, timedelta
from timeframe import fetch_financial_data

def get_times():
    """Get current time in different financial centers"""
    # Define the time zones
    greenwich_tz = pytz.timezone('GMT')
    london_tz = pytz.timezone('Europe/London')
    newyork_tz = pytz.timezone('America/New_York')
    tokyo_tz = pytz.timezone('Asia/Tokyo')
    sydney_tz = pytz.timezone('Australia/Sydney')

    # Get current UTC time
    utc_now = datetime.now(pytz.utc)

    # Convert to each timezone
    greenwich_time = utc_now.astimezone(greenwich_tz)
    london_time = utc_now.astimezone(london_tz)
    newyork_time = utc_now.astimezone(newyork_tz)
    tokyo_time = utc_now.astimezone(tokyo_tz)
    sydney_time = utc_now.astimezone(sydney_tz)

    # Format the times
    time_format = "%Y-%m-%d %H:%M:%S %Z%z"

    times = {
        "Greenwich": greenwich_time.strftime(time_format),
        "London": london_time.strftime(time_format),
        "New York": newyork_time.strftime(time_format),
        "Tokyo": tokyo_time.strftime(time_format),
        "Sydney": sydney_time.strftime(time_format)
    }

    return times

def chat_with_ai(chat_history, temperature=0.7):
    """Send chat history to AI and get response"""
    url = "https://corvo-ai-xx-ry.hf.space/chat"

    headers = {
        "Content-Type": "application/json",
    }

    payload = {
        "chat_history": chat_history,
        "temperature": temperature,
    }

    max_retries = 3
    retry_delay = 5

    for attempt in range(max_retries):
        try:
            response = requests.post(url, headers=headers, json=payload, timeout=120)
            response.raise_for_status()
            result = response.json()
            return result.get("assistant_response", "No response received.")
        except Exception as e:
            if attempt < max_retries - 1:
                time.sleep(retry_delay)

    return "عذراً، واجهت مشكلة في الاتصال. يرجى المحاولة مرة أخرى لاحقاً. 🙁"

def analyze_forex_pairs(pairs, deal_details):
    """Main function to analyze a list of forex pairs with deal details"""
    print(f"Analyzing forex pairs: {', '.join(pairs)}")

    # Step 1: Get current times in financial centers
    market_times = get_times()
    times_text = "\n".join([f"{k}: {v}" for k, v in market_times.items()])

    # Prepare system prompt with market times, deal details and group context
    system_prompt = f"""
🎯 Role: Professional Technical Assistant for Trade Monitoring and Cross-Pair Relationship Analysis  
You do **not** open trades. The trade is initiated by the Lead Analyst, or in some cases, the user is **waiting for a confirmation to enter** (e.g., 146.42 after a candle closes above a zone). Your job is to:

- Monitor any open trade carefully, or watch the price closely until the entry condition is met
- Analyze the market structure continuously using **both 5MIN and 15MIN timeframes**
- Study and assess dynamic relationships between correlated pairs and group momentum
- Alert the user **if a session (e.g., New York or London)** is about to close (within 15 minutes), as this may drastically impact market momentum
- Send at least one message for **each** pair
- Close the trade immediately if it becomes technically invalid

---
📌 Trade Details:
- 📋 Trade Information: {deal_details}\n\n
- 🕰️ Timeframes: {times_text}\n\n
- 📊 Currency Pairs: {', '.join(pairs)}

---
📌 Detailed Instructions:
1. ✅ Analyze the market using **5MIN and 15MIN charts**
   - General Trend (Up / Down / Sideways)
   - Market Structure: Is it still valid? Broken?
   - Momentum: Use RSI and MACD (Divergence? Weakness? Strength?)
   - Candle Patterns: Rejections, Engulfing, Fakeouts
   - Volume Activity: Any sudden surge or drop?
   - Session Impact (London, New York, etc.) — also check if the session is **nearing its end**
   - 📈 Correlation Analysis: Is the pair moving in sync with related pairs? Is there divergence? Is Group Momentum supporting the trade?
   - 🧠 **Review related pair charts and market dashboards** to understand the overall market sentiment and how it might impact the current trade.

---
📌 You **must always** return a structured technical analysis in this format:
<Analysis>
A clean, professional breakdown of the current trade situation based on 5M and 15M. Include trend, structure, momentum, correlation insights, session timing impact, and any critical zones.
</Analysis>

---
📬 You **must** send at least one user-facing message for each pair:
<send_message>
<message>
💬 A technical message in natural Arabic (or user’s native language) describing the current state of the trade or waiting condition.
- 📍 **Entry Price:** Use the value from the trade data
- 📊 **Pips Gained or Waiting:** Calculate the difference from the current price to entry, or note if price hasn’t reached the zone yet
- ✅ Current Technical Status: e.g., “Sideways range with weakening momentum,” or “Structure remains intact with support from 15M RSI,” or “Waiting for a confirmation candle on 5M”
</message>
<reason>
🧠 Explain **why** the alert is sent — use valid technical reasons like divergence, fake breakouts, weak candles, or session nearing its end.
</reason>
<Prediction>
Balanced forecast: Is the trade still strong? Are we waiting for confirmation? Is there a reversal risk? Also mention if an upcoming session end could affect things.
</Prediction>
<Advice>
💡 A clear, natural, and **non-repetitive** recommendation. Be expressive:
- "Keep monitoring, trend is stable for now" ✅
- "Momentum is weakening as the session nears its end — stay alert" ⚠️
- "Consider securing partial profits after +18 pips" 💰
- "Wait for confirmation on 5M before engaging" 🔍
- "Adjusting stop might be wise due to sudden volatility" ⛑️
</Advice>
</send_message>

---
🚨 If the trade becomes technically invalid, close it immediately with no hesitation:
<close_deal>
📉 Market structure is completely broken + strong reversal signals are present (e.g., clear rejection candles on 15M, negative divergence, session nearing close, and weakening momentum). Closing the trade is recommended to avoid unnecessary losses.
</close_deal>

---
📌 Core Rules:
- ❗ You **must send at least one message per pair** — silence is not allowed
- ❗ If the trade is clearly invalid → use <close_deal> and nothing else
- ❗ Only suggest “Secure Profits” if real profit exceeds 15 pips
- ❗ Avoid repetitive or templated advice — always be context-aware and natural
- ❗ Never use `<send_message>` and `<close_deal>` in the same response

---
🧠 Pro-Level Enhancements:
- ✅ Monitor alignment between correlated pairs
- ⚠️ Divergence + weak candles = Reversal risk
- ✅ Strong confirmation candle + rising volume = High-confidence continuation
- 🔄 If waiting for an entry condition (e.g., candle close), track price proximity and confirm status
- 💡 Use Gold, S&P500, and Dollar Index correlations to understand connected pair behavior
- ⏰ Warn the user if **a major session will close soon** — that can change market flow
- 📊 Always review multiple pair dashboards to forecast what's likely to happen to the current trade

---
🧘 You are a **professional assistant**:
- No emotions
- No guessing
- No unnecessary updates
- You speak **only when** there is a valid, professional-level technical insight
🎯 If the trade is valid → analyze and reassure
❌ If the trade is weak or invalid → close it immediately
🕵️ If waiting to enter → monitor until the technical entry condition is confirmed
"""


    # Build chat history with system prompt
    chat_history = [
        {"role": "system", "content": system_prompt}
    ]

    # Get timeframe data for each pair and add to chat history
    for pair in pairs:
        # Prepare the symbol for Yahoo Finance API
        symbol = pair
        if pair not in ["DX-Y.NYB", "GC=F", "WTI"]:
            symbol = f"{pair}=X"

        # Get 15min data for the last 3 days
        now = datetime.now()
        start_date_15m = now - timedelta(days=3)
        result_15m = fetch_financial_data(symbol=symbol, start_date=start_date_15m, end_date=now, interval="15m")

        # Get 5min data for the last 1 hour
        start_date_5m = now - timedelta(hours=1)
        result_5m = fetch_financial_data(symbol=symbol, start_date=start_date_5m, end_date=now, interval="5m")

        # Add the 15min data to chat history
        if result_15m['success']:
            chat_history.append({
                "role": "user",
                "content": f"TRADE USER STARTING \n\n{deal_details}\n\n## {pair} - 15 Minute Timeframe (Last 3 Days)\n\n### Meta Information\n{result_15m['meta_info']}\n\n{result_15m['table']}\n\n### Statistics\n{result_15m['stats']}"
            })

        # Add the 5min data to chat history
        if result_5m['success']:
            chat_history.append({
                "role": "user",
                "content": f"TRADE USER STARTING \n\n{deal_details}\n\n## {pair} - 5 Minute Timeframe (Last 1 Hour)\n\n### Meta Information\n{result_5m['meta_info']}\n\n{result_5m['table']}\n\n### Statistics\n{result_5m['stats']}"
            })

    # Get AI analysis
    analysis = chat_with_ai(chat_history)
    print(analysis)

    return analysis