Platform
    Our platform is a SaaS platform where many traders will use it for:
    a. Market study and help in learning trading
    b. Studying their own past transactions
    c. Interactive chat to learn more about market analysis or their transactions
Zerodha Connections
    Global
        a. We have a global Zerodha connection (Admin account) used for:
            Instruments
            Market data
            Historical candles
            Intraday candles
        b. Every day, Admin must log in and connect to Zerodha to activate the access token
    Personal
        a. Each user can connect their personal Zerodha account
        b. This account is used only to fetch transactions for analysis
        c. No trading or market data fetching is done using personal accounts
1. Watchlist (Daily)
    We have multiple lists of stocks for daily watchlist for intraday:
    a. Personal Watchlist
        User-specific stocks added manually
        Users manage their own watchlist
        This list flushes daily as bcoz, everyday new interest
    b. ET lists
        ET list: taken from ET movers API  or crawling
        From this list, we take:
        Top 10 gainers
        Top 10 losers

    c. Early Breakout / Breakdown
        This list is finalized before market hours
        Runs about 1 hour before market opens
        Uses only historical data
        Scans all ~3203 stocks from stocks_history
        Produces a short list of stocks ready for:
        Early breakout
        Early breakdown

2. Background Tasks
    a. Stock History Maintenance (Once in a day around 7 PM)
        Stocks masters collection - Create / update once in every 6 months 
        We run background tasks to keep stock history up to date:
        Timeframes stored:
        1 Day → max 300 candles or 1 year
        Initial storage:
            Fetch history for all stocks
        Ongoing updates:
            Fetch only missing or new data today's day by EOD
            Respect Zerodha API limits
            Partial updates are allowed
    b. Early movers (830 Pm Every day - Sunday to Thursday)
        Scans all ~3203 stocks from stocks_history
        Computes and finalize top 50 for potential early movers by algorithm
        This list is finalized before market hours
        Runs about 1 hour before market opens
        Uses only historical data
        This list is ready for the next day trade.
    
    c. Intraday Background Jobs
        We run background jobs for intraday data:
        Timeframes:
            5 min
            15 min
            30 min
        Data characteristics:
            Data is kept for current trading day & Last Trading day
            Data is flushed at EOD
            Intraday data is not stored long-term
        Lists covered:
            Personal Watchlist
            ET
            Early Breakout
    c. Scheduling Rules
        Early Breakout:
            Every 5 min from 9:00–10:00 AM
            After 10:00 AM → every 15 min (Using the same loop as Intra day - ET)
        ET list:
            Every 15 min
        personal watchlist:
            Every 15 min
            Except for list and time, the analysis logic is the same for all
3. Background Analysis Flow (Runs Every 15 Minutes)
    a. Build Global List
        Get lists from database
        Aggregate:
            Aggregated Personal watchlist (all users)
            ET list
            Early Movers list (only for 9 to 10 every 5 minutes)
            Remove duplicates
    b. Portfolio Rule
        No forced shorting
        Only deduplication
        Rest of the analysis remains same
    c. ET List Processing
        Take top 10 gainers & losers
    e. Indicator & Strategy Computation
        Strategy is intraday BUY or SELL
        No directional bias (LONG + SHORT both allowed)
        Compute:
            Indicators (6)
            Strategies (14)
            Timeframes used:
            5 min
            15 min
            30 min
            1 Day
            1 Week
            1 Month
    f. GPT Analysis
        GPT is called only for the filtered top list
        GPT is NOT called for the entire universe
    g. Data Storage
        Store:
        Snapshot data
        Computed indicators & strategies
        GPT output
    h. Stream
        Frontend fetches latest data using APIs
4. Chat
Chat is used in Stream, Alerts, and Learning.
    a. Analysis Chat
        Chat starts with latest snapshot (intraday) data
        Computed data passed:
        Indicators (6)
        Strategies (14)
        Goal:
        Get GPT input to help decide BUY / SELL / HOLD
    b. Transactions Chat
        Pass transaction details related to a trade
        Goal:
        Analyze whether the trade was right or wrong
        Additional data may be attached to improve analysis
5. Paper Trades
    a. Paper trades are automatically created when a potential BUY / SELL signal is ready
    b. Analysis score must be ≥ 80 (HIGH or VERY_STRONG)
    c. Virtual capital assigned: ₹10,00,000
    d. With every new data update, paper trades are created or updated
    e. Paper trade settings are controlled from backend
    f. Paper trades are closed at EOD for daily summary
    g. Account balance is updated based on profit or loss
6. Learning
    a. Users can learn from real trade transactions fetched from Zerodha
7. Alerts
    a. Alerts are produced only for:
    VERY_STRONG conviction
    Score ≥ 85
8. GPT Usage
    a. GPT is used only at specific analysis times
    b. GPT is used for learning transactions
    c. No unwanted or unnecessary GPT calls
    d. GPT is called only in background analysis
    e. Chat GPT calls are async
    f. GPT or Zerodha calls must not block any other process