delete stale remote cache folders

backend/.dockerignore

@@ -1,5 +0,0 @@
-__pycache__/
-*.pyc
-.venv/
-venv/
-.env

backend/Dockerfile

@@ -1,15 +0,0 @@
-FROM python:3.11-slim
-
-ENV PYTHONDONTWRITEBYTECODE=1
-ENV PYTHONUNBUFFERED=1
-
-WORKDIR /app
-
-COPY requirements.txt .
-RUN pip install --no-cache-dir -r requirements.txt
-
-COPY . .
-
-EXPOSE 7860
-
-CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860", "--ws", "none"]

backend/README.md

@@ -1,39 +0,0 @@
----
-title: NIFTY 50 Forecaster Backend
-emoji: 📈
-colorFrom: green
-colorTo: blue
-sdk: docker
-app_port: 7860
----
-
-# NIFTY 50 Forecaster Backend
-
-FastAPI Hugging Face Docker Space for the NIFTY 50 first-five-minute direction forecaster.
-
-## Endpoints
-
-- `GET /health`
-- `GET /dashboard`
-- `GET /prediction/latest`
-- `POST /prediction/refresh-first5`
-- `POST /data/refresh-daily`
-- `GET /cron/keepalive`
-- `POST /data/refresh-market-close`
-
-## Data
-
-Parquet files live in `data/`:
-
-- `nifty50_1m.parquet`
-- `nifty50_1d.parquet`
-- `opening_direction_training_dataset.parquet`
-- `test_predictions.parquet`
-
-## Runtime
-
-The API starts a daily background refresh loop. It wakes after `09:20 Asia/Kolkata`, fetches Yahoo Finance `^NSEI` 1-minute candles for the `09:15-09:19` opening window, appends them to Parquet, and writes the latest T+5 prediction.
-
-After market close it wakes again at `15:45 Asia/Kolkata`, refreshes the 1-minute and daily Parquet files, updates the opening training dataset with same-day close outcomes, and writes the saved prediction record used for the next trading session card. The `/cron/keepalive` endpoint also checks this close refresh so a Hugging Face Space that was idled still catches up when Netlify pings it.
-
-Netlify also pings `/cron/keepalive` every 10 minutes through its scheduled function.

backend/__init__.py

@@ -1,2 +0,0 @@
-"""NIFTY Project backend package."""
-

backend/app.py

@@ -1,536 +0,0 @@
-from __future__ import annotations
-
-import asyncio
-import threading
-from datetime import date, datetime, time, timedelta
-
-import sys
-from pathlib import Path
-
-from fastapi import BackgroundTasks, HTTPException, Query
-from fastapi.middleware.cors import CORSMiddleware
-from fastapi import FastAPI
-from pydantic import BaseModel
-
-sys.path.insert(0, str(Path(__file__).resolve().parent))
-from nifty_backend.runtime import (
-    CLOSE_REFRESH_READY,
-    IST,
-    STALE_CHECK_INTERVAL_SECONDS,
-    TPLUS1_READY,
-    close_refresh_due,
-    dashboard_payload,
-    is_trading_day,
-    latest_saved_prediction,
-    latest_tplus1_prediction,
-    next_trading_day,
-    refresh_daily_data,
-    refresh_first5_prediction,
-    refresh_market_close_data,
-    refresh_stale_data_once,
-    refresh_tplus1_prediction,
-    seconds_until_next_ist_run,
-    warm_dashboard_payload_cache,
-)
-from kotak_neo import (
-    KotakNeoConfigError,
-    KotakNeoError,
-    KotakNeoSessionRequired,
-    kotak_neo_manager,
-)
-
-
-app = FastAPI(title="NIFTY 50 Forecaster Backend")
-app.add_middleware(
-    CORSMiddleware,
-    allow_origins=["*"],
-    allow_credentials=False,
-    allow_methods=["*"],
-    allow_headers=["*"],
-)
-
-
-market_status = "Waiting for next session"
-close_refresh_lock = threading.Lock()
-tplus1_refresh_lock = threading.Lock()
-MARKET_OPEN = time(9, 15)
-FIRST5_READY = time(9, 20)
-MARKET_CLOSE = time(15, 30)
-
-
-class TotpRequest(BaseModel):
-    totp: str
-
-
-def refresh_market_close_data_if_due() -> dict:
-    if not close_refresh_due():
-        return {"status": "skipped", "reason": "close refresh is not due"}
-    if not close_refresh_lock.acquire(blocking=False):
-        return {"status": "skipped", "reason": "close refresh already running"}
-    try:
-        info = refresh_market_close_data()
-        return {"status": "refreshed", **info}
-    finally:
-        close_refresh_lock.release()
-
-
-def latest_tplus1_prediction_date(payload: dict | None = None) -> date | None:
-    try:
-        latest = payload if payload is not None else latest_tplus1_prediction()
-        raw = latest.get("input_date")
-        return date.fromisoformat(str(raw)[:10]) if raw else None
-    except Exception:
-        return None
-
-
-def tplus1_refresh_due(now: datetime | None = None, latest_date: date | None = None) -> bool:
-    now = now or datetime.now(IST)
-    if not is_trading_day(now.date()) or not (TPLUS1_READY <= now.time() < MARKET_CLOSE):
-        return False
-    latest_date = latest_date if latest_date is not None else latest_tplus1_prediction_date()
-    return latest_date != now.date()
-
-
-def refresh_tplus1_if_due() -> dict:
-    now = datetime.now(IST)
-    latest_date = latest_tplus1_prediction_date()
-    if not tplus1_refresh_due(now=now, latest_date=latest_date):
-        return {"status": "skipped", "reason": "tplus1 refresh is not due"}
-    if not tplus1_refresh_lock.acquire(blocking=False):
-        return {"status": "skipped", "reason": "tplus1 refresh already running"}
-    try:
-        prediction = refresh_tplus1_prediction(session_date=now.date())
-        return {"status": "refreshed", "prediction": prediction}
-    finally:
-        tplus1_refresh_lock.release()
-
-
-def latest_prediction_date(payload: dict | None = None) -> date | None:
-    try:
-        latest = payload if payload is not None else latest_saved_prediction()
-        raw = latest.get("input_date")
-        return date.fromisoformat(str(raw)) if raw else None
-    except Exception:
-        return None
-
-
-def current_market_state(now: datetime | None = None) -> dict:
-    global market_status
-    now = now or datetime.now(IST)
-    today = now.date()
-    current_time = now.time()
-    trading_day = is_trading_day(today)
-    latest_date = latest_prediction_date()
-    market_is_open_for_t5 = trading_day and FIRST5_READY <= current_time < MARKET_CLOSE
-    market_is_open_for_tplus1 = trading_day and TPLUS1_READY <= current_time < MARKET_CLOSE
-    has_current_first5 = market_is_open_for_t5 and latest_date == today
-    tplus1_latest_date = latest_tplus1_prediction_date()
-    has_current_tplus1 = market_is_open_for_tplus1 and tplus1_latest_date == today
-    next_session = today if trading_day and current_time < MARKET_CLOSE else next_trading_day(today + timedelta(days=1))
-
-    if not trading_day:
-        status = "Market Closed"
-        detail = f"Next trading session is {next_session.isoformat()}."
-    elif current_time < time(9, 0):
-        status = "Waiting for 9:00 AM"
-        detail = "Market has not entered pre-open yet."
-    elif current_time < MARKET_OPEN:
-        status = "Market Pre-Open"
-        detail = "Market opens at 9:15 AM IST."
-    elif current_time < FIRST5_READY:
-        status = "Market Officially Opened"
-        detail = "Waiting for the first 5 one-minute bars."
-    elif current_time <= MARKET_CLOSE:
-        if market_status in {"Fetching T+5 Prediction Data...", "Prediction Failed"}:
-            status = market_status
-            detail = "The first-five-minute prediction job is still resolving."
-        elif has_current_first5:
-            status = "Prediction Ready"
-            detail = "Today's first-five-minute prediction is available."
-        else:
-            status = "Prediction Pending"
-            detail = "No current-session prediction has been generated yet."
-    else:
-        status = "Market Closed"
-        detail = "Trading session has ended."
-
-    if not trading_day:
-        tplus1_status = "Market Closed"
-        tplus1_detail = f"Next trading session is {next_session.isoformat()}."
-    elif current_time < TPLUS1_READY:
-        tplus1_status = "Waiting for 2:30 PM"
-        tplus1_detail = "The T+1 forecast becomes available at 2:30 PM IST."
-    elif current_time < MARKET_CLOSE:
-        if has_current_tplus1:
-            tplus1_status = "Ready"
-            tplus1_detail = "Today's T+1 prediction is available."
-        else:
-            tplus1_status = "Pending"
-            tplus1_detail = "No current-session T+1 prediction has been generated yet."
-    else:
-        tplus1_status = "Market Closed"
-        tplus1_detail = "Trading session has ended."
-
-    if not trading_day:
-        t5_status = "Market Closed"
-        t5_detail = f"Next trading session is {next_session.isoformat()}."
-    elif current_time < FIRST5_READY:
-        t5_status = "Waiting for 9:20 AM"
-        t5_detail = "The T+5 forecast becomes available after the first five one-minute bars."
-    elif current_time < MARKET_CLOSE:
-        if market_status in {"Fetching T+5 Prediction Data...", "Prediction Failed"}:
-            t5_status = market_status
-            t5_detail = "The first-five-minute prediction job is still resolving."
-        elif has_current_first5:
-            t5_status = "Ready"
-            t5_detail = "Today's first-five-minute prediction is available."
-        else:
-            t5_status = "Pending"
-            t5_detail = "No current-session prediction has been generated yet."
-    else:
-        t5_status = "Market Closed"
-        t5_detail = "Trading session has ended."
-
-    return {
-        "market_status": status,
-        "market_detail": detail,
-        "server_time_ist": now.isoformat(),
-        "is_trading_day": trading_day,
-        "session_date": today.isoformat(),
-        "next_session_date": next_session.isoformat(),
-        "latest_prediction_date": latest_date.isoformat() if latest_date else None,
-        "t5_available": has_current_first5,
-        "t5_status": t5_status,
-        "t5_detail": t5_detail,
-        "market_is_open_for_t5": market_is_open_for_t5,
-        "tplus1_available": has_current_tplus1,
-        "tplus1_status": tplus1_status,
-        "tplus1_detail": tplus1_detail,
-        "market_is_open_for_tplus1": market_is_open_for_tplus1,
-        "latest_tplus1_prediction_date": tplus1_latest_date.isoformat() if tplus1_latest_date else None,
-    }
-
-
-def attach_market_state(payload: dict) -> dict:
-    state = current_market_state()
-    payload.setdefault("data_status", {})
-    payload["data_status"].update(state)
-    try:
-        payload["nifty_quote"] = kotak_neo_manager.fetch_nifty50_quote()
-        payload["nifty_quote_error"] = None
-    except KotakNeoSessionRequired as exc:
-        payload["nifty_quote"] = None
-        payload["nifty_quote_error"] = {"status": 401, "message": str(exc)}
-    except KotakNeoConfigError as exc:
-        payload["nifty_quote"] = None
-        payload["nifty_quote_error"] = {"status": 503, "message": str(exc)}
-    except KotakNeoError as exc:
-        payload["nifty_quote"] = None
-        payload["nifty_quote_error"] = {"status": 502, "message": str(exc)}
-
-    t5_latest = payload.get("latest") or {}
-    tomorrow_latest = payload.get("tomorrow_latest") or {}
-    tplus1_latest = payload.get("tplus1_latest") or {}
-    t5_available = bool(state["t5_available"] and t5_latest.get("prediction"))
-    tplus1_available = bool(state["tplus1_available"] and tplus1_latest.get("prediction"))
-    tomorrow_available = bool(tomorrow_latest.get("prediction"))
-    refresh_phase = payload.get("data_status", {}).get("refresh_phase")
-    if refresh_phase in {"waiting_second_payload", "refreshing"}:
-        tomorrow_status = "WAITING FOR SECOND PAYLOAD"
-        tomorrow_reason = "Market close refresh is generating the next-session payload."
-    else:
-        tomorrow_status = "Ready" if tomorrow_available else "Pending"
-        tomorrow_reason = None if tomorrow_available else "No saved next-session signal is available."
-    payload["predictions"] = {
-        "tomorrow": {
-            "available": tomorrow_available,
-            "status": tomorrow_status,
-            "reason": tomorrow_reason,
-            "target_date": tomorrow_latest.get("target_date") or state["next_session_date"],
-            "input_date": tomorrow_latest.get("input_date"),
-            "prediction": tomorrow_latest.get("prediction") if tomorrow_available else None,
-            "prob_up": tomorrow_latest.get("prob_up") if tomorrow_available else None,
-            "confidence": tomorrow_latest.get("confidence") if tomorrow_available else None,
-            "threshold": tomorrow_latest.get("threshold") if tomorrow_available else None,
-            "model_name": tomorrow_latest.get("model_name"),
-            "source_model": tomorrow_latest.get("source_model"),
-            "validation_accuracy": tomorrow_latest.get("validation_accuracy"),
-            "test_accuracy": tomorrow_latest.get("test_accuracy"),
-        },
-        "t5": {
-            "available": t5_available,
-            "status": "Ready" if t5_available else state["t5_status"],
-            "reason": None if t5_available else state["t5_detail"],
-            "input_date": t5_latest.get("input_date"),
-            "prediction": t5_latest.get("prediction") if t5_available else None,
-            "prob_up": t5_latest.get("prob_up") if t5_available else None,
-            "confidence": t5_latest.get("confidence") if t5_available else None,
-            "threshold": t5_latest.get("threshold") if t5_available else None,
-            "model_name": t5_latest.get("model_name"),
-            "validation_accuracy": (payload.get("summary") or {}).get("validation_accuracy"),
-            "test_accuracy": (payload.get("summary") or {}).get("test_accuracy"),
-        },
-        "tplus1": {
-            "available": tplus1_available,
-            "status": "Ready" if tplus1_available else state["tplus1_status"],
-            "reason": None if tplus1_available else state["tplus1_detail"],
-            "target_date": tplus1_latest.get("target_date") or state["next_session_date"],
-            "input_date": tplus1_latest.get("input_date"),
-            "prediction": tplus1_latest.get("prediction") if tplus1_available else None,
-            "prob_up": tplus1_latest.get("prob_up") if tplus1_available else None,
-            "confidence": tplus1_latest.get("confidence") if tplus1_available else None,
-            "threshold": tplus1_latest.get("threshold") if tplus1_available else None,
-            "model_name": tplus1_latest.get("model_name"),
-            "validation_accuracy": (payload.get("tplus1_summary") or {}).get("validation_accuracy"),
-            "test_accuracy": (payload.get("tplus1_summary") or {}).get("test_accuracy"),
-        },
-    }
-    return payload
-
-
-async def daily_ist_refresh_loop() -> None:
-    global market_status
-    while True:
-        # Wait until 9:00 AM IST
-        await asyncio.sleep(seconds_until_next_ist_run(time(9, 0)))
-        if not is_trading_day(datetime.now(IST).date()):
-            market_status = "Market Closed"
-            continue
-        market_status = "Market Pre-Open"
-        print("[scheduler] 9:00 AM IST - Market Pre-Open", flush=True)
-
-        # Wait until 9:15 AM IST
-        await asyncio.sleep(seconds_until_next_ist_run(time(9, 15)))
-        market_status = "Market Officially Opened"
-        print("[scheduler] 9:15 AM IST - Market Officially Opened", flush=True)
-
-        # Wait until 9:20 AM IST
-        await asyncio.sleep(seconds_until_next_ist_run(time(9, 20)))
-        market_status = "Fetching T+5 Prediction Data..."
-        print("[scheduler] 9:20 AM IST - Fetching Data", flush=True)
-
-        try:
-            await asyncio.to_thread(refresh_first5_prediction)
-            market_status = "Prediction Ready"
-        except Exception as exc:
-            print(f"[scheduler] first5 refresh failed: {exc}", flush=True)
-            market_status = "Prediction Failed"
-            
-        try:
-            await asyncio.to_thread(refresh_daily_data)
-        except Exception as exc:
-            print(f"[scheduler] daily refresh failed: {exc}", flush=True)
-
-        await asyncio.sleep(seconds_until_next_ist_run(TPLUS1_READY))
-        print("[scheduler] 2:30 PM IST - Refreshing T+1 prediction", flush=True)
-        try:
-            info = await asyncio.to_thread(refresh_tplus1_if_due)
-            print(f"[scheduler] tplus1 refresh result: {info}", flush=True)
-        except Exception as exc:
-            print(f"[scheduler] tplus1 refresh failed: {exc}", flush=True)
-
-        await asyncio.sleep(seconds_until_next_ist_run(CLOSE_REFRESH_READY))
-        print("[scheduler] 3:45 PM IST - Refreshing close data", flush=True)
-        try:
-            info = await asyncio.to_thread(refresh_market_close_data_if_due)
-            print(f"[scheduler] close refresh result: {info}", flush=True)
-        except Exception as exc:
-            print(f"[scheduler] close refresh failed: {exc}", flush=True)
-
-
-async def refresh_current_session_once() -> None:
-    global market_status
-    now = datetime.now(IST)
-    if not is_trading_day(now.date()) or now.time() < FIRST5_READY:
-        return
-    if latest_prediction_date() == now.date():
-        return
-    market_status = "Fetching T+5 Prediction Data..."
-    print("[startup] Current session needs first-five refresh; fetching now.", flush=True)
-    try:
-        await asyncio.to_thread(refresh_first5_prediction)
-        market_status = "Prediction Ready"
-    except Exception as exc:
-        print(f"[startup] first5 refresh failed: {exc}", flush=True)
-        market_status = "Prediction Failed"
-    try:
-        await asyncio.to_thread(refresh_daily_data)
-    except Exception as exc:
-        print(f"[startup] daily refresh failed: {exc}", flush=True)
-
-
-async def refresh_market_close_once_if_due() -> None:
-    try:
-        info = await asyncio.to_thread(refresh_market_close_data_if_due)
-        if info.get("status") == "refreshed":
-            print(f"[startup] close refresh result: {info}", flush=True)
-    except Exception as exc:
-        print(f"[startup] close refresh failed: {exc}", flush=True)
-
-
-async def refresh_tplus1_once_if_due() -> None:
-    try:
-        info = await asyncio.to_thread(refresh_tplus1_if_due)
-        if info.get("status") == "refreshed":
-            print(f"[startup] tplus1 refresh result: {info}", flush=True)
-    except Exception as exc:
-        print(f"[startup] tplus1 refresh failed: {exc}", flush=True)
-
-
-async def warm_dashboard_payload_cache_once() -> None:
-    try:
-        await asyncio.to_thread(warm_dashboard_payload_cache)
-    except Exception as exc:
-        print(f"[startup] dashboard payload warmup failed: {exc}", flush=True)
-
-
-async def stale_data_watch_loop() -> None:
-    while True:
-        try:
-            info = await asyncio.to_thread(refresh_stale_data_once)
-            if info.get("status") == "refreshed":
-                print(f"[stale-watch] refreshed stale data: {info}", flush=True)
-        except Exception as exc:
-            print(f"[stale-watch] stale refresh failed: {exc}", flush=True)
-        await asyncio.sleep(STALE_CHECK_INTERVAL_SECONDS)
-
-
-@app.on_event("startup")
-async def start_scheduler() -> None:
-    global market_status
-    # Initialize correct status on startup based on current time
-    now = datetime.now(IST).time()
-    today = datetime.now(IST).date()
-    if not is_trading_day(today):
-        market_status = "Market Closed"
-    elif now < time(9, 0):
-        market_status = "Waiting for 9:00 AM"
-    elif now < time(9, 15):
-        market_status = "Market Pre-Open"
-    elif now < time(9, 20):
-        market_status = "Market Officially Opened"
-    elif latest_prediction_date() == today:
-        market_status = "Prediction Ready"
-    else:
-        market_status = "Prediction Pending"
-        
-    asyncio.create_task(refresh_current_session_once())
-    asyncio.create_task(refresh_tplus1_once_if_due())
-    asyncio.create_task(refresh_market_close_once_if_due())
-    asyncio.create_task(warm_dashboard_payload_cache_once())
-    asyncio.create_task(stale_data_watch_loop())
-    asyncio.create_task(daily_ist_refresh_loop())
-
-
-@app.get("/health")
-def health() -> dict[str, str]:
-    return {"status": "ok"}
-
-
-@app.get("/")
-def root() -> dict[str, str]:
-    return {"service": "NIFTY 50 Forecaster Backend", "status": "ok"}
-
-
-@app.get("/dashboard")
-def dashboard() -> dict:
-    return attach_market_state(dashboard_payload())
-
-
-@app.get("/kotak/status")
-def kotak_status() -> dict:
-    return kotak_neo_manager.status()
-
-
-@app.post("/kotak/auth/totp")
-def kotak_auth_totp(payload: TotpRequest) -> dict:
-    try:
-        return kotak_neo_manager.authenticate_with_totp(payload.totp)
-    except KotakNeoConfigError as exc:
-        raise HTTPException(status_code=503, detail=str(exc)) from exc
-    except KotakNeoError as exc:
-        raise HTTPException(status_code=400, detail=str(exc)) from exc
-
-
-@app.get("/kotak/account")
-def kotak_account() -> dict:
-    try:
-        return kotak_neo_manager.fetch_account_snapshot()
-    except KotakNeoConfigError as exc:
-        raise HTTPException(status_code=503, detail=str(exc)) from exc
-    except KotakNeoSessionRequired as exc:
-        raise HTTPException(status_code=401, detail=str(exc)) from exc
-    except KotakNeoError as exc:
-        raise HTTPException(status_code=502, detail=str(exc)) from exc
-
-
-@app.get("/kotak/quote/nifty50")
-def kotak_nifty50_quote() -> dict:
-    try:
-        return kotak_neo_manager.fetch_nifty50_quote()
-    except KotakNeoConfigError as exc:
-        raise HTTPException(status_code=503, detail=str(exc)) from exc
-    except KotakNeoSessionRequired as exc:
-        raise HTTPException(status_code=401, detail=str(exc)) from exc
-    except KotakNeoError as exc:
-        raise HTTPException(status_code=502, detail=str(exc)) from exc
-
-
-@app.get("/kotak/activity-log")
-def kotak_activity_log() -> dict:
-    try:
-        snapshot = kotak_neo_manager.fetch_account_snapshot()
-        return {
-            "activity_log": snapshot.get("activity_log", {}),
-            "trade_history": snapshot.get("trade_history", []),
-            "order_book": snapshot.get("order_book", []),
-        }
-    except KotakNeoConfigError as exc:
-        raise HTTPException(status_code=503, detail=str(exc)) from exc
-    except KotakNeoSessionRequired as exc:
-        raise HTTPException(status_code=401, detail=str(exc)) from exc
-    except KotakNeoError as exc:
-        raise HTTPException(status_code=502, detail=str(exc)) from exc
-
-
-@app.get("/cron/keepalive")
-def cron_keepalive(background_tasks: BackgroundTasks) -> dict:
-    close_refresh = {"status": "not_checked"}
-    tplus1_refresh = {"status": "not_checked"}
-    if tplus1_refresh_due():
-        background_tasks.add_task(refresh_tplus1_if_due)
-        tplus1_refresh = {"status": "scheduled"}
-    if close_refresh_due():
-        background_tasks.add_task(refresh_market_close_data_if_due)
-        close_refresh = {"status": "scheduled"}
-    return {
-        "status": "awake",
-        "market": current_market_state(),
-        "tplus1_refresh": tplus1_refresh,
-        "close_refresh": close_refresh,
-    }
-
-
-@app.get("/prediction/latest")
-def prediction_latest() -> dict:
-    return latest_saved_prediction()
-
-
-@app.post("/prediction/refresh-first5")
-def prediction_refresh_first5(
-    session_date: date | None = Query(default=None, description="Optional YYYY-MM-DD session date in IST."),
-) -> dict:
-    prediction = refresh_first5_prediction(session_date=session_date)
-    return prediction.to_dict()
-
-
-@app.post("/data/refresh-daily")
-def data_refresh_daily() -> dict:
-    return refresh_daily_data()
-
-
-@app.post("/data/refresh-market-close")
-def data_refresh_market_close(
-    session_date: date | None = Query(default=None, description="Optional YYYY-MM-DD session date in IST."),
-) -> dict:
-    return refresh_market_close_data(session_date=session_date)

backend/data/nifty50_1d.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:4d0830f9d3cfa91f02ce717b556983beb12f897bc537a623c67e38f22ce05caf
-size 78366

backend/data/nifty50_1m.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:b5d6022df273daa1f020676214ec35536426fd84f2d7efb669797287e27ffa2d
-size 18589635

backend/data/opening_direction_training_dataset.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:722c8225451abcb49463d2a57354bc0c9b5eb519f31a25fa3c5242adc4dbbada
-size 4463631

backend/data/test_predictions.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:f159a7499394b7882262ffaa6f9b48c0f6ab763d024f373ee19109b301af90c1
-size 14499

backend/data/tomorrow_test_predictions.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:a2b6ecb377d114c825f861d9a6741e01bbf23bfc623493fd01c78e8bb1501960
-size 10751

backend/data/tplus1_test_predictions.parquet

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:bd7f63239d5705969cbe5423169c9fdcb88d59b838b02913d0aa5c455a7ca41c
-size 13681

backend/kotak_neo.py

@@ -1,1257 +0,0 @@
-from __future__ import annotations
-
-import os
-import threading
-from csv import DictReader
-from concurrent.futures import ThreadPoolExecutor, as_completed
-from datetime import date, datetime, time, timezone
-from functools import lru_cache
-from pathlib import Path
-from time import monotonic
-from typing import Any
-from urllib.parse import quote
-from zoneinfo import ZoneInfo
-
-import json
-
-import pandas as pd
-import requests
-
-try:
-    import pandas_market_calendars as mcal
-except Exception:  # pragma: no cover - deployed environments may fall back to weekdays
-    mcal = None
-
-
-SESSION_BASE_URL = "https://mis.kotaksecurities.com"
-QUOTE_PATH_TEMPLATE = "script-details/1.0/quotes/neosymbol/{neo_symbols}/{quote_type}"
-TOTP_LOGIN_PATH = "login/1.0/tradeApiLogin"
-TOTP_VALIDATE_PATH = "login/1.0/tradeApiValidate"
-DEFAULT_TIMEOUT_SECONDS = 20
-ACCOUNT_TIMEOUT_SECONDS = 7
-NIFTY_QUOTE_TIMEOUT_SECONDS = 2.5
-NIFTY_QUOTE_CACHE_SECONDS = 1.0
-DATA_DIR = Path(__file__).resolve().parent / "data"
-KOTAK_ACTIVITY_LOG_PATH = DATA_DIR / "kotak_activity_log.txt"
-NIFTY_1M_PATH = DATA_DIR / "nifty50_1m.parquet"
-NIFTY_1D_PATH = DATA_DIR / "nifty50_1d.parquet"
-IST = ZoneInfo("Asia/Kolkata")
-MARKET_OPEN_TIME = time(9, 15)
-MARKET_CLOSE_TIME = time(15, 30)
-
-
-class KotakNeoError(Exception):
-    pass
-
-
-class KotakNeoConfigError(KotakNeoError):
-    pass
-
-
-class KotakNeoSessionRequired(KotakNeoError):
-    pass
-
-
-def _utc_now_iso() -> str:
-    return datetime.now(timezone.utc).isoformat()
-
-
-def _to_float(value: Any) -> float | None:
-    if value in (None, "", "--", "NA", "na", "-", "null"):
-        return None
-    try:
-        return float(value)
-    except (TypeError, ValueError):
-        return None
-
-
-def _first_number(*values: Any) -> float | None:
-    for value in values:
-        parsed = _to_float(value)
-        if parsed is not None:
-            return parsed
-    return None
-
-
-def _first_market_number(*values: Any) -> float | None:
-    fallback = None
-    for value in values:
-        parsed = _to_float(value)
-        if parsed is None:
-            continue
-        if fallback is None:
-            fallback = parsed
-        if parsed > 0:
-            return parsed
-    return fallback
-
-
-def _normalize_frame_dates(frame: pd.DataFrame) -> pd.Series:
-    values = pd.to_datetime(frame["date"], errors="coerce")
-    if getattr(values.dt, "tz", None) is None:
-        values = values.dt.tz_localize(IST)
-    else:
-        values = values.dt.tz_convert(IST)
-    return values
-
-
-@lru_cache(maxsize=1)
-def _nse_calendar():
-    if mcal is None:
-        return None
-    for name in ("XNSE", "NSE", "BSE"):
-        try:
-            return mcal.get_calendar(name)
-        except Exception:
-            continue
-    return None
-
-
-@lru_cache(maxsize=64)
-def _is_nse_trading_day(day: date) -> bool:
-    calendar = _nse_calendar()
-    if calendar is None:
-        return day.weekday() < 5
-    return not calendar.schedule(start_date=day, end_date=day).empty
-
-
-@lru_cache(maxsize=64)
-def _previous_nse_trading_day(day: date) -> date:
-    candidate = day
-    for _ in range(21):
-        candidate = date.fromordinal(candidate.toordinal() - 1)
-        if _is_nse_trading_day(candidate):
-            return candidate
-    return candidate
-
-
-def _file_version(path: Path) -> tuple[str, int | None, int | None]:
-    try:
-        stat = path.stat()
-        return (str(path), stat.st_mtime_ns, stat.st_size)
-    except OSError:
-        return (str(path), None, None)
-
-
-@lru_cache(maxsize=4)
-def _load_nifty_daily_frame(file_version: tuple[str, int | None, int | None]) -> pd.DataFrame:
-    path = Path(file_version[0])
-    daily = pd.read_parquet(path, columns=["date", "open", "high", "low", "close"]).copy()
-    daily["date"] = pd.to_datetime(daily["date"], errors="coerce").dt.date
-    return daily.dropna(subset=["date"]).sort_values("date")
-
-
-@lru_cache(maxsize=4)
-def _load_nifty_minute_frame(file_version: tuple[str, int | None, int | None]) -> pd.DataFrame:
-    path = Path(file_version[0])
-    minute = pd.read_parquet(path, columns=["date", "open", "high", "low", "close"]).copy()
-    minute["date"] = _normalize_frame_dates(minute)
-    return minute.dropna(subset=["date"]).sort_values("date")
-
-
-def _first_text(*values: Any) -> str | None:
-    for value in values:
-        if isinstance(value, dict):
-            nested = _first_text(
-                value.get("message"),
-                value.get("error"),
-                value.get("Error"),
-                value.get("emsg"),
-                value.get("detail"),
-            )
-            if nested:
-                return nested
-        if isinstance(value, list):
-            for item in value:
-                nested = _first_text(item)
-                if nested:
-                    return nested
-        if value not in (None, "", "--", "NA", "na", "-"):
-            return str(value)
-    return None
-
-
-def _sum_numbers(*values: Any) -> float | None:
-    numbers: list[float] = []
-    for value in values:
-        parsed = _to_float(value)
-        if parsed is not None:
-            numbers.append(parsed)
-    return sum(numbers) if numbers else None
-
-
-def _extract_items(payload: Any) -> list[dict[str, Any]]:
-    if isinstance(payload, list):
-        return [item for item in payload if isinstance(item, dict)]
-    if not isinstance(payload, dict):
-        return []
-
-    data = payload.get("data")
-    if isinstance(data, list):
-        return [item for item in data if isinstance(item, dict)]
-    if isinstance(data, dict):
-        nested = data.get("data")
-        if isinstance(nested, list):
-            return [item for item in nested if isinstance(item, dict)]
-        return [data]
-    return []
-
-
-def _sort_key(item: dict[str, Any]) -> str:
-    return str(
-        _first_text(
-            item.get("updRecvTm"),
-            item.get("hsUpTm"),
-            item.get("flDtTm"),
-            item.get("exTm"),
-            item.get("ordDtTm"),
-            item.get("TimeStamp"),
-            item.get("flDt"),
-        )
-        or ""
-    )
-
-
-class KotakNeoManager:
-    def __init__(self) -> None:
-        self.consumer_key = os.getenv("KOTAK_CONSUMER_KEY")
-        self.mobile_number = os.getenv("KOTAK_MOBILE_NUMBER")
-        self.ucc = os.getenv("KOTAK_UCC")
-        self.mpin = os.getenv("KOTAK_MPIN")
-        self.neo_fin_key = os.getenv("KOTAK_NEO_FIN_KEY", "neotradeapi")
-
-        self._lock = threading.RLock()
-        self.activity_log_path = KOTAK_ACTIVITY_LOG_PATH
-        self.activity_log_path.parent.mkdir(parents=True, exist_ok=True)
-        self._seen_activity_keys: set[str] = set()
-        self._scrip_cache: dict[str, list[dict[str, str]]] = {}
-        self._quote_cache: dict[str, dict[str, Any]] = {}
-        self._load_existing_activity_keys()
-        self._clear_session_locked()
-
-    def _load_existing_activity_keys(self) -> None:
-        if not self.activity_log_path.exists():
-            return
-        try:
-            for line in self.activity_log_path.read_text(encoding="utf-8").splitlines():
-                if not line.strip():
-                    continue
-                try:
-                    payload = json.loads(line)
-                except json.JSONDecodeError:
-                    continue
-                key = str(payload.get("activity_key") or "").strip()
-                if key:
-                    self._seen_activity_keys.add(key)
-        except Exception:
-            pass
-
-    def _clear_session_locked(self) -> None:
-        self.view_token: str | None = None
-        self.sid: str | None = None
-        self.edit_token: str | None = None
-        self.edit_sid: str | None = None
-        self.edit_rid: str | None = None
-        self.server_id: str | None = None
-        self.data_center: str | None = None
-        self.base_url: str | None = None
-        self.authenticated_at: str | None = None
-
-    def _configured(self) -> bool:
-        return all([self.consumer_key, self.mobile_number, self.ucc, self.mpin])
-
-    def status(self) -> dict[str, Any]:
-        configured = self._configured()
-        with self._lock:
-            authenticated = bool(self.edit_token and self.edit_sid and self.base_url)
-            return {
-                "available": configured,
-                "configured": configured,
-                "authenticated": authenticated,
-                "needs_totp": configured and not authenticated,
-                "last_authenticated_at": self.authenticated_at,
-                "reason": None if configured else "Kotak Neo environment variables are incomplete.",
-            }
-
-    def authenticate_with_totp(self, totp: str) -> dict[str, Any]:
-        if not self._configured():
-            raise KotakNeoConfigError("Kotak Neo environment variables are incomplete.")
-        if not str(totp).strip():
-            raise KotakNeoError("A TOTP code is required.")
-
-        with self._lock:
-            login_response = self._post_session_api(
-                TOTP_LOGIN_PATH,
-                headers={
-                    "Authorization": self.consumer_key,
-                    "neo-fin-key": self.neo_fin_key,
-                    "Content-Type": "application/json",
-                    "Accept": "application/json",
-                },
-                payload={
-                    "mobileNumber": self.mobile_number,
-                    "ucc": self.ucc,
-                    "totp": str(totp).strip(),
-                },
-            )
-            login_data = (login_response.get("data") or {}) if isinstance(login_response, dict) else {}
-            self.view_token = login_data.get("token")
-            self.sid = login_data.get("sid")
-
-            if not self.view_token or not self.sid:
-                self._clear_session_locked()
-                raise KotakNeoError("Kotak Neo did not return a valid pre-auth session.")
-
-            validate_response = self._post_session_api(
-                TOTP_VALIDATE_PATH,
-                headers={
-                    "Authorization": self.consumer_key,
-                    "sid": self.sid,
-                    "Auth": self.view_token,
-                    "neo-fin-key": self.neo_fin_key,
-                    "Content-Type": "application/json",
-                    "Accept": "application/json",
-                },
-                payload={"mpin": self.mpin},
-            )
-            validate_data = (validate_response.get("data") or {}) if isinstance(validate_response, dict) else {}
-
-            self.edit_token = validate_data.get("token")
-            self.edit_sid = validate_data.get("sid")
-            self.edit_rid = validate_data.get("rid")
-            self.server_id = validate_data.get("hsServerId")
-            self.data_center = validate_data.get("dataCenter")
-            self.base_url = str(validate_data.get("baseUrl") or "").rstrip("/")
-            self.authenticated_at = _utc_now_iso()
-
-            if not self.edit_token or not self.edit_sid or not self.base_url:
-                self._clear_session_locked()
-                raise KotakNeoError("Kotak Neo did not return a usable trading session.")
-
-            return self.status()
-
-    def fetch_account_snapshot(self) -> dict[str, Any]:
-        if not self._configured():
-            raise KotakNeoConfigError("Kotak Neo environment variables are incomplete.")
-
-        with self._lock:
-            context = self._context_locked()
-
-        account_calls = {
-            "holdings": lambda: self._request_trading_api_with_context(
-                context,
-                "portfolio/v1/holdings",
-                timeout=ACCOUNT_TIMEOUT_SECONDS,
-            ),
-            "positions": lambda: self._request_trading_api_with_context(
-                context,
-                "quick/user/positions",
-                timeout=ACCOUNT_TIMEOUT_SECONDS,
-            ),
-            "trades": lambda: self._request_trading_api_with_context(
-                context,
-                "quick/user/trades",
-                timeout=ACCOUNT_TIMEOUT_SECONDS,
-            ),
-            "orders": lambda: self._request_trading_api_with_context(
-                context,
-                "quick/user/orders",
-                timeout=ACCOUNT_TIMEOUT_SECONDS,
-            ),
-            "limits": lambda: self._post_trading_api_with_context(
-                context,
-                "quick/user/limits",
-                payload={"seg": "ALL", "exch": "ALL", "prod": "ALL"},
-                content_type="application/x-www-form-urlencoded",
-                timeout=ACCOUNT_TIMEOUT_SECONDS,
-            ),
-        }
-        defaults = {
-            "holdings": {"data": []},
-            "positions": {"data": []},
-            "trades": {"data": []},
-            "orders": {"data": []},
-            "limits": {},
-        }
-        results: dict[str, dict[str, Any]] = dict(defaults)
-
-        with ThreadPoolExecutor(max_workers=5) as executor:
-            future_map = {executor.submit(fn): label for label, fn in account_calls.items()}
-            for future in as_completed(future_map):
-                label = future_map[future]
-                results[label] = self._resolve_account_future(label, future, default=defaults[label])
-
-        holdings = _extract_items(results["holdings"])
-        positions = _extract_items(results["positions"])
-        trades = sorted(_extract_items(results["trades"]), key=_sort_key, reverse=True)
-        orders = sorted(_extract_items(results["orders"]), key=_sort_key, reverse=True)
-
-        normalized_trades = [self._normalize_trade(item) for item in trades]
-        normalized_orders = [self._normalize_order(item) for item in orders]
-
-        quotes = self._safe_account_call(
-            "quotes",
-            lambda: self._fetch_quotes_with_context(
-                context,
-                self._instrument_tokens_for_quotes(
-                    context,
-                    holdings,
-                    positions,
-                    normalized_trades,
-                ),
-                timeout=ACCOUNT_TIMEOUT_SECONDS,
-            )
-            ,
-            default={"data": []},
-        )
-        quote_map = self._build_quote_map(quotes)
-
-        normalized_holdings = [self._normalize_holding(item, quote_map) for item in holdings]
-        normalized_positions = [self._normalize_position(item, quote_map) for item in positions]
-        self._append_activity_entries(normalized_trades, normalized_orders)
-        journal = self._read_activity_journal()
-        merged_trades = self._merge_activity(normalized_trades, journal["trades"])
-        merged_orders = self._merge_activity(normalized_orders, journal["orders"])
-
-        holdings_market_value = sum(item["market_value"] or 0.0 for item in normalized_holdings)
-        holdings_cost = sum(item["cost_value"] or 0.0 for item in normalized_holdings)
-        holdings_pnl = sum(item["pnl"] or 0.0 for item in normalized_holdings)
-        positions_pnl = sum(item["pnl"] or 0.0 for item in normalized_positions)
-
-        limits_raw = results["limits"]
-        limits_summary = {
-            "net": _first_number(limits_raw.get("Net")) if isinstance(limits_raw, dict) else None,
-            "margin_used": _first_number(limits_raw.get("MarginUsed")) if isinstance(limits_raw, dict) else None,
-            "collateral_value": _first_number(limits_raw.get("CollateralValue")) if isinstance(limits_raw, dict) else None,
-            "cash_unrealized_mtm": _first_number(limits_raw.get("CashUnRlsMtomPrsnt")) if isinstance(limits_raw, dict) else None,
-            "cash_realized_mtm": _first_number(limits_raw.get("CashRlsMtomPrsnt")) if isinstance(limits_raw, dict) else None,
-        }
-
-        available_cash = None
-        if limits_summary["net"] is not None and limits_summary["margin_used"] is not None:
-            available_cash = limits_summary["net"] - limits_summary["margin_used"]
-
-        current_capital = None
-        if available_cash is not None:
-            current_capital = available_cash + holdings_market_value
-
-        return {
-            "status": self.status(),
-            "as_of": _utc_now_iso(),
-            "neo_behavior": {
-                "holdings_note": "Kotak Neo shows CNC delivery buys in Positions on trade day and in Holdings/T1 from the next trading day.",
-                "trade_history_note": "Kotak Neo trade history availability is limited by Neo's own order and portfolio tracker behavior.",
-            },
-            "summary": {
-                "available_cash": available_cash,
-                "current_capital": current_capital,
-                "holdings_market_value": holdings_market_value,
-                "holdings_cost_value": holdings_cost,
-                "holdings_pnl": holdings_pnl,
-                "positions_pnl": positions_pnl,
-                "live_pnl": holdings_pnl + positions_pnl,
-                "open_positions": sum(1 for item in normalized_positions if item["net_quantity"]),
-                "holdings_count": len(normalized_holdings),
-                "orders_count": len(merged_orders),
-                "trades_count": len(merged_trades),
-            },
-            "limits_summary": limits_summary,
-            "limits_raw": limits_raw,
-            "holdings": normalized_holdings,
-            "positions": normalized_positions,
-            "trade_history": merged_trades[:100],
-            "order_book": merged_orders[:100],
-            "activity_log": {
-                "path": str(self.activity_log_path),
-                "trades_count": len(journal["trades"]),
-                "orders_count": len(journal["orders"]),
-            },
-            "quotes": list(quote_map.values()),
-        }
-
-    def fetch_nifty50_quote(self, *, force_refresh: bool = False) -> dict[str, Any]:
-        cache_key = "nifty50_quote"
-        with self._lock:
-            if not force_refresh:
-                cached = self._quote_cache.get(cache_key)
-                if cached and (monotonic() - float(cached.get("stored_at_monotonic") or 0.0)) < NIFTY_QUOTE_CACHE_SECONDS:
-                    return dict(cached["payload"])
-            context = self._context_locked()
-
-        reference = self._resolve_nifty50_reference(context)
-        quote_payload = self._fetch_quotes_with_context(
-            context,
-            [
-                {
-                    "exchange_segment": "nse_cm",
-                    "instrument_token": str(reference["quote_instrument_token"]),
-                }
-            ],
-            timeout=NIFTY_QUOTE_TIMEOUT_SECONDS,
-        )
-        items = _extract_items(quote_payload)
-        if not items:
-            raise KotakNeoError("Kotak Neo did not return a NIFTY 50 quote.")
-
-        item = items[0]
-        now_ist = datetime.now(IST)
-        last_traded_price = _first_market_number(item.get("last_traded_price"), item.get("ltp"), item.get("iv"))
-        quote_open = _first_market_number(item.get("openingPrice"), item.get("open"), item.get("o"))
-        quote_high = _first_market_number(item.get("high"), item.get("highPrice"), item.get("h"))
-        quote_low = _first_market_number(item.get("low"), item.get("lowPrice"), item.get("l"))
-        quote_previous_close = _first_market_number(
-            item.get("previous_close"),
-            item.get("previousClose"),
-            item.get("prev_close"),
-            item.get("prevClose"),
-            item.get("previousClosePrice"),
-            item.get("prevClosePrice"),
-            item.get("close"),
-            item.get("c"),
-            item.get("ic"),
-        )
-        live_stats = self._load_nifty50_reference_stats(
-            now_ist,
-            last_traded_price,
-            quote_open,
-            quote_high,
-            quote_low,
-            quote_previous_close,
-        )
-        close = live_stats["previous_close"]
-        change_base = live_stats["return_base"]
-        change = None
-        change_pct = None
-        if last_traded_price is not None and change_base not in (None, 0):
-            change = last_traded_price - change_base
-            change_pct = (change / change_base) * 100.0
-        high = live_stats["range_high"]
-        low = live_stats["range_low"]
-        open_price = quote_open
-
-        payload = {
-            "symbol": "NIFTY 50",
-            "exchange_segment": "nse_cm",
-            "instrument_token": _first_text(
-                item.get("instrument_token"),
-                item.get("instrumentToken"),
-                item.get("tk"),
-                reference.get("master_instrument_token"),
-                reference["quote_instrument_token"],
-            ),
-            "display_name": _first_text(item.get("trading_symbol"), item.get("ts"), item.get("name"), "NIFTY 50"),
-            "last_traded_price": last_traded_price,
-            "close": close,
-            "change": change,
-            "change_pct": change_pct,
-            "open": open_price,
-            "high": high,
-            "low": low,
-            "return_basis": live_stats["return_basis"],
-            "market_open": live_stats["market_open"],
-            "is_trading_session": live_stats["is_trading_session"],
-            "quote_session_date": live_stats["quote_session_date"],
-            "previous_session_date": live_stats["previous_session_date"],
-            "exchange_feed_time": _first_text(item.get("tvalue"), item.get("updRecvTm"), item.get("hsUpTm")),
-            "as_of": _utc_now_iso(),
-            "source": {
-                "quote_api": QUOTE_PATH_TEMPLATE,
-                "master_scrip_verified": bool(reference.get("master_record_found")),
-                "instrument_lookup": reference.get("lookup_mode"),
-                "master_symbol_name": reference.get("master_symbol_name"),
-                "master_trading_symbol": reference.get("master_trading_symbol"),
-                "reference_data": "backend/data/nifty50_1m.parquet + backend/data/nifty50_1d.parquet",
-            },
-        }
-
-        with self._lock:
-            self._quote_cache[cache_key] = {
-                "stored_at_monotonic": monotonic(),
-                "payload": payload,
-            }
-        return payload
-
-    def _load_nifty50_reference_stats(
-        self,
-        now_ist: datetime,
-        last_traded_price: float | None,
-        live_open: float | None = None,
-        live_high: float | None = None,
-        live_low: float | None = None,
-        live_previous_close: float | None = None,
-    ) -> dict[str, Any]:
-        today = now_ist.date()
-        is_trading_session = _is_nse_trading_day(today)
-        market_open = is_trading_session and MARKET_OPEN_TIME <= now_ist.time() < MARKET_CLOSE_TIME
-        session_started = is_trading_session and now_ist.time() >= MARKET_OPEN_TIME
-        quote_session_date = today if session_started else _previous_nse_trading_day(today)
-        previous_session_date = _previous_nse_trading_day(quote_session_date)
-
-        daily = _load_nifty_daily_frame(_file_version(NIFTY_1D_PATH))
-
-        previous_close = None
-        session_daily = daily[daily["date"] == quote_session_date]
-        previous_daily = daily[daily["date"] == previous_session_date]
-        if previous_daily.empty:
-            previous_daily = daily[daily["date"] < quote_session_date]
-        if not previous_daily.empty:
-            previous_row = previous_daily.iloc[-1]
-            previous_close = _to_float(previous_row["close"])
-
-        session_open = _to_float(session_daily.iloc[-1]["open"]) if not session_daily.empty else None
-        session_high = _to_float(session_daily.iloc[-1]["high"]) if not session_daily.empty else None
-        session_low = _to_float(session_daily.iloc[-1]["low"]) if not session_daily.empty else None
-        session_close = _to_float(session_daily.iloc[-1]["close"]) if not session_daily.empty else None
-
-        if live_previous_close is not None:
-            is_same_as_static_ltp = (
-                last_traded_price is not None
-                and abs(live_previous_close - last_traded_price) < 0.01
-            )
-            if not is_same_as_static_ltp:
-                previous_close = live_previous_close
-
-        if session_started:
-            session_open = live_open or session_open
-            session_high = live_high or session_high
-            session_low = live_low or session_low
-
-            if session_open is None or session_high is None or session_low is None:
-                minute = _load_nifty_minute_frame(_file_version(NIFTY_1M_PATH))
-                today_minute = minute[minute["date"].dt.date == today]
-                if not today_minute.empty:
-                    session_open = _to_float(today_minute.iloc[0]["open"]) or session_open
-                    minute_high = pd.to_numeric(today_minute["high"], errors="coerce").max()
-                    minute_low = pd.to_numeric(today_minute["low"], errors="coerce").min()
-                    session_high = _to_float(minute_high) or session_high
-                    session_low = _to_float(minute_low) or session_low
-                    session_close = _to_float(today_minute.iloc[-1]["close"]) or session_close
-        else:
-            session_high = live_high or session_high
-            session_low = live_low or session_low
-
-        if session_started:
-            range_high = max([value for value in [session_high, last_traded_price] if value is not None], default=None)
-            range_low = min([value for value in [session_low, last_traded_price] if value is not None], default=None)
-            return_base = previous_close
-            return_basis = "previous_close"
-        else:
-            range_high = session_high
-            range_low = session_low
-            return_base = previous_close
-            return_basis = "previous_close"
-            if last_traded_price is None:
-                last_traded_price = session_close
-
-        return {
-            "previous_close": previous_close,
-            "return_base": return_base,
-            "return_basis": return_basis,
-            "range_high": range_high,
-            "range_low": range_low,
-            "market_open": market_open,
-            "is_trading_session": is_trading_session,
-            "quote_session_date": quote_session_date.isoformat(),
-            "previous_session_date": previous_session_date.isoformat(),
-        }
-
-    def _ensure_authenticated_locked(self) -> None:
-        if not self.edit_token or not self.edit_sid or not self.base_url:
-            raise KotakNeoSessionRequired("Kotak Neo session is not authenticated.")
-
-    def _context_locked(self) -> dict[str, str]:
-        self._ensure_authenticated_locked()
-        return {
-            "base_url": self.base_url or "",
-            "edit_sid": self.edit_sid or "",
-            "edit_token": self.edit_token or "",
-            "server_id": self.server_id or "",
-            "consumer_key": self.consumer_key or "",
-        }
-
-    def _post_session_api(self, path: str, headers: dict[str, str], payload: dict[str, Any]) -> dict[str, Any]:
-        response = requests.post(
-            f"{SESSION_BASE_URL.rstrip('/')}/{path.lstrip('/')}",
-            headers=headers,
-            json=payload,
-            timeout=DEFAULT_TIMEOUT_SECONDS,
-        )
-        data = self._decode_response(response)
-        self._raise_for_error(response, data, session_sensitive=False)
-        return data
-
-    def _request_trading_api_locked(self, path: str) -> dict[str, Any]:
-        return self._request_trading_api_with_context(
-            self._context_locked(),
-            path,
-            timeout=DEFAULT_TIMEOUT_SECONDS,
-        )
-
-    def _request_trading_api_with_context(
-        self,
-        context: dict[str, str],
-        path: str,
-        *,
-        timeout: int,
-    ) -> dict[str, Any]:
-        response = requests.get(
-            f"{context['base_url'].rstrip('/')}/{path.lstrip('/')}",
-            headers={
-                "Sid": context["edit_sid"],
-                "Auth": context["edit_token"],
-                "Accept": "application/json",
-            },
-            params={"sId": context["server_id"]},
-            timeout=timeout,
-        )
-        data = self._decode_response(response)
-        self._raise_for_error(response, data, session_sensitive=True)
-        return data
-
-    def _post_trading_api_locked(
-        self,
-        path: str,
-        payload: dict[str, Any],
-        *,
-        content_type: str = "application/json",
-    ) -> dict[str, Any]:
-        return self._post_trading_api_with_context(
-            self._context_locked(),
-            path,
-            payload,
-            content_type=content_type,
-            timeout=DEFAULT_TIMEOUT_SECONDS,
-        )
-
-    def _post_trading_api_with_context(
-        self,
-        context: dict[str, str],
-        path: str,
-        payload: dict[str, Any],
-        *,
-        content_type: str,
-        timeout: int,
-    ) -> dict[str, Any]:
-        headers = {
-            "Sid": context["edit_sid"],
-            "Auth": context["edit_token"],
-            "Accept": "application/json",
-            "Content-Type": content_type,
-        }
-        query_params = {"sId": context["server_id"]}
-        url = f"{context['base_url'].rstrip('/')}/{path.lstrip('/')}"
-
-        if content_type == "application/x-www-form-urlencoded":
-            body = {"jData": json.dumps(payload)}
-            response = requests.post(
-                url,
-                headers=headers,
-                params=query_params,
-                data=body,
-                timeout=timeout,
-            )
-        else:
-            response = requests.post(
-                url,
-                headers=headers,
-                params=query_params,
-                json=payload,
-                timeout=timeout,
-            )
-
-        data = self._decode_response(response)
-        self._raise_for_error(response, data, session_sensitive=True)
-        return data
-
-    def _fetch_quotes_locked(self, instrument_tokens: list[dict[str, str]]) -> dict[str, Any]:
-        return self._fetch_quotes_with_context(
-            self._context_locked(),
-            instrument_tokens,
-            timeout=DEFAULT_TIMEOUT_SECONDS,
-        )
-
-    def _fetch_quotes_with_context(
-        self,
-        context: dict[str, str],
-        instrument_tokens: list[dict[str, str]],
-        *,
-        timeout: int,
-    ) -> dict[str, Any]:
-        if not instrument_tokens:
-            return {"data": []}
-
-        neo_symbols = ",".join(
-            f"{item['exchange_segment']}|{item['instrument_token']}" for item in instrument_tokens
-        )
-        encoded_symbols = quote(neo_symbols, safe="")
-        response = requests.get(
-            f"{context['base_url'].rstrip('/')}/{QUOTE_PATH_TEMPLATE.format(neo_symbols=encoded_symbols, quote_type='all')}",
-            headers={
-                "Authorization": context["consumer_key"],
-                "Content-Type": "application/x-www-form-urlencoded",
-                "Accept": "application/json",
-            },
-            timeout=timeout,
-        )
-        data = self._decode_response(response)
-        self._raise_for_error(response, data, session_sensitive=False)
-        return data
-
-    def _decode_response(self, response: requests.Response) -> dict[str, Any]:
-        try:
-            parsed = response.json()
-        except ValueError as exc:
-            raise KotakNeoError(
-                f"Kotak Neo returned a non-JSON response with HTTP {response.status_code}."
-            ) from exc
-        if isinstance(parsed, dict):
-            return parsed
-        return {"data": parsed}
-
-    def _safe_account_call(
-        self,
-        label: str,
-        fn,
-        *,
-        default: dict[str, Any],
-    ) -> dict[str, Any]:
-        try:
-            return fn()
-        except KotakNeoSessionRequired:
-            raise
-        except Exception as exc:
-            if not self._is_expected_empty_error(label, exc):
-                print(f"[kotak] {label} call failed: {exc}", flush=True)
-            return default
-
-    def _resolve_account_future(self, label: str, future, *, default: dict[str, Any]) -> dict[str, Any]:
-        try:
-            return future.result()
-        except KotakNeoSessionRequired:
-            raise
-        except Exception as exc:
-            if not self._is_expected_empty_error(label, exc):
-                print(f"[kotak] {label} call failed: {exc}", flush=True)
-            return default
-
-    def _is_expected_empty_error(self, label: str, exc: Exception) -> bool:
-        text = str(exc).lower()
-        empty_markers = [
-            "no holdings found",
-            "no position",
-            "no positions",
-            "no trade",
-            "no trades",
-            "no order",
-            "no orders",
-            "no data found",
-            "no trade found",
-            "no order found",
-        ]
-        if any(marker in text for marker in empty_markers):
-            return True
-        if label in {"positions", "trades", "orders"} and text.strip() == "kotak neo rejected the request.":
-            return True
-        if label in {"holdings", "positions", "trades", "orders"} and "424" in text:
-            return True
-        return False
-
-    def _raise_for_error(
-        self,
-        response: requests.Response,
-        data: dict[str, Any],
-        *,
-        session_sensitive: bool,
-    ) -> None:
-        stat = str(data.get("stat") or "").strip().lower()
-        st_code = _first_number(data.get("stCode"))
-        message = _first_text(
-            data.get("message"),
-            data.get("error"),
-            data.get("Error"),
-            data.get("emsg"),
-        )
-
-        session_expired = response.status_code in (401, 403) or "invalid session" in (message or "").lower()
-        if session_expired:
-            if session_sensitive:
-                self._clear_session_locked()
-                raise KotakNeoSessionRequired(message or "Kotak Neo session expired.")
-            raise KotakNeoError(message or "Kotak Neo rejected the request.")
-
-        if stat == "not_ok":
-            raise KotakNeoError(message or "Kotak Neo rejected the request.")
-
-        if response.status_code >= 400 or (st_code is not None and st_code >= 400):
-            raise KotakNeoError(message or f"Kotak Neo request failed with HTTP {response.status_code}.")
-
-    def _instrument_tokens_for_quotes(
-        self,
-        context: dict[str, str],
-        holdings: list[dict[str, Any]],
-        positions: list[dict[str, Any]],
-        trades: list[dict[str, Any]] | None = None,
-    ) -> list[dict[str, str]]:
-        unique: dict[tuple[str, str], dict[str, str]] = {}
-
-        for item in holdings:
-            token = _first_text(item.get("instrumentToken"), item.get("exchangeIdentifier"), item.get("tok"))
-            exchange = _first_text(item.get("exchangeSegment"), item.get("exSeg"))
-            if token and exchange:
-                unique[(exchange, token)] = {
-                    "exchange_segment": exchange,
-                    "instrument_token": token,
-                }
-
-        for item in positions:
-            token = _first_text(item.get("tok"), item.get("instrumentToken"), item.get("exchangeIdentifier"))
-            exchange = _first_text(item.get("exSeg"), item.get("exchangeSegment"))
-            if token and exchange:
-                unique[(exchange, token)] = {
-                    "exchange_segment": exchange,
-                    "instrument_token": token,
-                }
-            elif exchange:
-                resolved = self._resolve_symbol_token(
-                    context,
-                    exchange_segment=exchange,
-                    symbol=_first_text(item.get("trdSym"), item.get("sym")),
-                )
-                if resolved:
-                    unique[(exchange, resolved)] = {
-                        "exchange_segment": exchange,
-                        "instrument_token": resolved,
-                    }
-
-        for item in trades or []:
-            exchange = _first_text(item.get("exchange_segment"), item.get("exSeg"))
-            if not exchange:
-                continue
-            resolved = self._resolve_symbol_token(
-                context,
-                exchange_segment=exchange,
-                symbol=_first_text(item.get("trading_symbol"), item.get("trdSym"), item.get("symbol"), item.get("sym")),
-            )
-            if resolved:
-                unique[(exchange, resolved)] = {
-                    "exchange_segment": exchange,
-                    "instrument_token": resolved,
-                }
-
-        return list(unique.values())
-
-    def _resolve_symbol_token(
-        self,
-        context: dict[str, str],
-        *,
-        exchange_segment: str,
-        symbol: str | None,
-    ) -> str | None:
-        symbol = str(symbol or "").strip()
-        if not symbol:
-            return None
-        candidates = self._load_scrip_candidates(context, exchange_segment)
-        symbol_upper = symbol.upper()
-        base_symbol_upper = symbol_upper.split("-")[0]
-        for item in candidates:
-            trading_symbol = str(item.get("pTrdSymbol") or "").upper()
-            symbol_name = str(item.get("pSymbolName") or "").upper()
-            token = str(item.get("pSymbol") or "").strip()
-            if not token:
-                continue
-            if trading_symbol == symbol_upper or symbol_name == base_symbol_upper:
-                return token
-        return None
-
-    def _load_scrip_candidates(self, context: dict[str, str], exchange_segment: str) -> list[dict[str, str]]:
-        key = str(exchange_segment).lower()
-        if key in self._scrip_cache:
-            return self._scrip_cache[key]
-
-        response = requests.get(
-            f"{context['base_url'].rstrip('/')}/script-details/1.0/masterscrip/file-paths",
-            headers={
-                "Authorization": context["consumer_key"],
-                "Accept": "application/json",
-            },
-            timeout=ACCOUNT_TIMEOUT_SECONDS,
-        )
-        data = self._decode_response(response)
-        file_paths = ((data.get("data") or {}).get("filesPaths") or []) if isinstance(data, dict) else []
-        csv_url = next((path for path in file_paths if key in str(path).lower()), None)
-        if not csv_url:
-            self._scrip_cache[key] = []
-            return []
-
-        csv_response = requests.get(csv_url, timeout=ACCOUNT_TIMEOUT_SECONDS)
-        csv_response.raise_for_status()
-        rows = list(DictReader(csv_response.text.splitlines()))
-        self._scrip_cache[key] = rows
-        return rows
-
-    def _resolve_nifty50_reference(self, context: dict[str, str]) -> dict[str, Any]:
-        candidates = self._load_scrip_candidates(context, "nse_cm")
-        match = None
-        aliases = {
-            "NIFTY 50",
-            "NIFTY50",
-            "NIFTY 50 INDEX",
-            "NIFTY",
-        }
-        for item in candidates:
-            candidate_values = {
-                str(item.get("pSymbolName") or "").strip().upper(),
-                str(item.get("pTrdSymbol") or "").strip().upper(),
-                str(item.get("pSymbol") or "").strip().upper(),
-            }
-            if aliases & candidate_values:
-                match = item
-                break
-
-        return {
-            "quote_instrument_token": "Nifty 50",
-            "lookup_mode": "index-name-direct" if match is None else "master-scrip-verified-index-name-direct",
-            "master_record_found": match is not None,
-            "master_instrument_token": _first_text(match.get("pSymbol")) if match else None,
-            "master_symbol_name": _first_text(match.get("pSymbolName")) if match else None,
-            "master_trading_symbol": _first_text(match.get("pTrdSymbol")) if match else None,
-        }
-
-    def _build_quote_map(self, payload: dict[str, Any]) -> dict[str, dict[str, Any]]:
-        items = _extract_items(payload)
-        quote_map: dict[str, dict[str, Any]] = {}
-        for item in items:
-            token = _first_text(item.get("instrument_token"), item.get("instrumentToken"), item.get("tk"))
-            exchange = _first_text(item.get("exchange_segment"), item.get("exchangeSegment"), item.get("e"))
-            if not token or not exchange:
-                continue
-            key = f"{exchange}|{token}"
-            quote_map[key] = {
-                "instrument_token": token,
-                "exchange_segment": exchange,
-                "trading_symbol": _first_text(item.get("trading_symbol"), item.get("ts"), item.get("symbol")),
-                "last_traded_price": _first_number(item.get("last_traded_price"), item.get("ltp"), item.get("iv")),
-                "close": _first_number(item.get("close"), item.get("c"), item.get("ic")),
-                "change": _first_number(item.get("change"), item.get("cng")),
-                "change_pct": _first_number(item.get("net_change_percentage"), item.get("nc")),
-            }
-        return quote_map
-
-    def _normalize_holding(
-        self,
-        item: dict[str, Any],
-        quote_map: dict[str, dict[str, Any]],
-    ) -> dict[str, Any]:
-        token = _first_text(item.get("instrumentToken"), item.get("exchangeIdentifier"), item.get("tok"))
-        exchange = _first_text(item.get("exchangeSegment"), item.get("exSeg"))
-        quote = quote_map.get(f"{exchange}|{token}", {}) if token and exchange else {}
-
-        quantity = _first_number(item.get("quantity"), item.get("sellableQuantity"))
-        average_price = _first_number(item.get("averagePrice"), item.get("avgPrc"))
-        holding_cost = _first_number(item.get("holdingCost"))
-        market_value = _first_number(item.get("mktValue"))
-        ltp = _first_number(quote.get("last_traded_price"))
-
-        if market_value is None and quantity is not None and ltp is not None:
-            market_value = quantity * ltp
-        if holding_cost is None and quantity is not None and average_price is not None:
-            holding_cost = quantity * average_price
-
-        pnl = None
-        pnl_pct = None
-        if market_value is not None and holding_cost is not None:
-            pnl = market_value - holding_cost
-            if holding_cost:
-                pnl_pct = pnl / holding_cost
-
-        return {
-            "symbol": _first_text(item.get("displaySymbol"), item.get("symbol"), item.get("trdSym")),
-            "trading_symbol": _first_text(item.get("symbol"), item.get("displaySymbol"), item.get("trdSym")),
-            "exchange_segment": exchange,
-            "instrument_token": token,
-            "quantity": quantity,
-            "sellable_quantity": _first_number(item.get("sellableQuantity")),
-            "average_price": average_price,
-            "last_traded_price": ltp,
-            "market_value": market_value,
-            "cost_value": holding_cost,
-            "pnl": pnl,
-            "pnl_pct": pnl_pct,
-        }
-
-    def _normalize_position(
-        self,
-        item: dict[str, Any],
-        quote_map: dict[str, dict[str, Any]],
-    ) -> dict[str, Any]:
-        token = _first_text(item.get("tok"), item.get("instrumentToken"), item.get("exchangeIdentifier"))
-        exchange = _first_text(item.get("exSeg"), item.get("exchangeSegment"))
-        quote = quote_map.get(f"{exchange}|{token}", {}) if token and exchange else {}
-
-        multiplier = _first_number(item.get("multiplier")) or 1.0
-        buy_qty = _first_number(
-            item.get("buyQty"),
-            _sum_numbers(item.get("cfBuyQty"), item.get("flBuyQty")),
-        )
-        sell_qty = _first_number(
-            item.get("sellQty"),
-            _sum_numbers(item.get("cfSellQty"), item.get("flSellQty")),
-        )
-        qty = _first_number(item.get("netQty"), item.get("qty"))
-
-        if qty is None and buy_qty is not None and sell_qty is not None:
-            qty = buy_qty - sell_qty
-        elif qty is None and buy_qty is not None and sell_qty is None:
-            qty = buy_qty
-        elif qty is None and sell_qty is not None:
-            qty = -sell_qty
-
-        average_price = _first_number(item.get("avgPrc"), item.get("averagePrice"))
-        ltp = _first_number(quote.get("last_traded_price"))
-
-        pnl = _first_number(
-            item.get("pnl"),
-            item.get("mtm"),
-            item.get("urmtom"),
-            item.get("unRealizedMtom"),
-        )
-        if pnl is None and qty is not None and average_price is not None and ltp is not None:
-            pnl = (ltp - average_price) * qty * multiplier
-
-        return {
-            "order_no": _first_text(item.get("nOrdNo")),
-            "symbol": _first_text(item.get("sym"), item.get("trdSym")),
-            "trading_symbol": _first_text(item.get("trdSym"), item.get("sym")),
-            "exchange_segment": exchange,
-            "instrument_token": token,
-            "product": _first_text(item.get("prod")),
-            "transaction_type": _first_text(item.get("trnsTp")),
-            "net_quantity": qty,
-            "average_price": average_price,
-            "last_traded_price": ltp,
-            "pnl": pnl,
-            "multiplier": multiplier,
-            "updated_at": _first_text(item.get("hsUpTm"), item.get("exTm"), item.get("flDtTm")),
-        }
-
-
-    def _normalize_trade(self, item: dict[str, Any]) -> dict[str, Any]:
-        return {
-            "activity_type": "trade",
-            "activity_key": self._trade_key(item),
-            "order_no": _first_text(item.get("nOrdNo")),
-            "trade_id": _first_text(item.get("flId")),
-            "exchange_order_id": _first_text(item.get("exOrdId")),
-            "symbol": _first_text(item.get("sym"), item.get("trdSym")),
-            "trading_symbol": _first_text(item.get("trdSym"), item.get("sym")),
-            "exchange_segment": _first_text(item.get("exSeg")),
-            "transaction_type": _first_text(item.get("trnsTp")),
-            "product": _first_text(item.get("prod")),
-            "quantity": _first_number(item.get("fldQty"), item.get("qty")),
-            "price": _first_number(item.get("avgPrc"), item.get("prc")),
-            "average_price": _first_number(item.get("avgPrc")),
-            "status": _first_text(item.get("rptTp"), item.get("ordSt"), item.get("stat")),
-            "trade_time": _first_text(item.get("flDtTm"), item.get("exTm"), item.get("flTm"), item.get("flDt")),
-            "raw": item,
-        }
-
-    def _normalize_order(self, item: dict[str, Any]) -> dict[str, Any]:
-        return {
-            "activity_type": "order",
-            "activity_key": self._order_key(item),
-            "order_no": _first_text(item.get("nOrdNo")),
-            "request_id": _first_text(item.get("reqId"), item.get("nReqId")),
-            "exchange_order_id": _first_text(item.get("exOrdId")),
-            "symbol": _first_text(item.get("sym"), item.get("trdSym")),
-            "trading_symbol": _first_text(item.get("trdSym"), item.get("sym")),
-            "exchange_segment": _first_text(item.get("exSeg")),
-            "transaction_type": _first_text(item.get("trnsTp")),
-            "product": _first_text(item.get("prod")),
-            "quantity": _first_number(item.get("qty")),
-            "filled_quantity": _first_number(item.get("fldQty")),
-            "unfilled_size": _first_number(item.get("unFldSz")),
-            "price": _first_number(item.get("prc")),
-            "trigger_price": _first_number(item.get("trgPrc")),
-            "order_type": _first_text(item.get("prcTp")),
-            "status": _first_text(item.get("ordSt"), item.get("stat")),
-            "order_time": _first_text(item.get("ordDtTm"), item.get("exCfmTm"), item.get("hsUpTm")),
-            "rejection_reason": _first_text(item.get("rejRsn")),
-            "raw": item,
-        }
-
-    def _trade_key(self, item: dict[str, Any]) -> str:
-        return "|".join(
-            [
-                "trade",
-                str(_first_text(item.get("nOrdNo")) or ""),
-                str(_first_text(item.get("flId")) or ""),
-                str(_first_text(item.get("flDtTm"), item.get("exTm"), item.get("flTm")) or ""),
-            ]
-        )
-
-    def _order_key(self, item: dict[str, Any]) -> str:
-        return "|".join(
-            [
-                "order",
-                str(_first_text(item.get("nOrdNo")) or ""),
-                str(_first_text(item.get("ordSt"), item.get("stat")) or ""),
-                str(_first_text(item.get("ordDtTm"), item.get("exCfmTm"), item.get("hsUpTm")) or ""),
-            ]
-        )
-
-    def _append_activity_entries(self, trades: list[dict[str, Any]], orders: list[dict[str, Any]]) -> None:
-        lines: list[str] = []
-        for entry in [*trades, *orders]:
-            key = str(entry.get("activity_key") or "").strip()
-            if not key or key in self._seen_activity_keys:
-                continue
-            payload = {
-                "activity_key": key,
-                "activity_type": entry.get("activity_type"),
-                "captured_at": _utc_now_iso(),
-                **entry,
-            }
-            lines.append(json.dumps(payload, ensure_ascii=True))
-            self._seen_activity_keys.add(key)
-        if not lines:
-            return
-        with self.activity_log_path.open("a", encoding="utf-8") as handle:
-            handle.write("\n".join(lines) + "\n")
-
-    def _read_activity_journal(self) -> dict[str, list[dict[str, Any]]]:
-        trades: list[dict[str, Any]] = []
-        orders: list[dict[str, Any]] = []
-        if not self.activity_log_path.exists():
-            return {"trades": trades, "orders": orders}
-        for line in self.activity_log_path.read_text(encoding="utf-8").splitlines():
-            if not line.strip():
-                continue
-            try:
-                item = json.loads(line)
-            except json.JSONDecodeError:
-                continue
-            if item.get("activity_type") == "trade":
-                trades.append(item)
-            elif item.get("activity_type") == "order":
-                orders.append(item)
-        trades.sort(key=lambda item: str(item.get("trade_time") or item.get("captured_at") or ""), reverse=True)
-        orders.sort(key=lambda item: str(item.get("order_time") or item.get("captured_at") or ""), reverse=True)
-        return {"trades": trades, "orders": orders}
-
-    def _merge_activity(
-        self,
-        live_entries: list[dict[str, Any]],
-        journal_entries: list[dict[str, Any]],
-    ) -> list[dict[str, Any]]:
-        merged: list[dict[str, Any]] = []
-        seen: set[str] = set()
-        for entry in [*live_entries, *journal_entries]:
-            key = str(entry.get("activity_key") or "").strip()
-            if not key or key in seen:
-                continue
-            merged.append(entry)
-            seen.add(key)
-        return merged
-
-
-kotak_neo_manager = KotakNeoManager()

backend/models/candidate_results.csv

@@ -1,14 +0,0 @@
-model_name,threshold,validation_accuracy,test_accuracy,validation_auc,test_auc
-blend_extra_trees_tight_logit_overlay,0.425,0.654320987654321,0.6524064171122995,0.6623500611995106,0.6563861499656042
-blend_extra_trees_tight_logit,0.425,0.6444444444444445,0.6310160427807486,0.6623500611995106,0.6563861499656042
-extra_trees_opening,0.514,0.6345679012345679,0.6203208556149733,0.6448470012239902,0.6575326759917449
-extra_trees_opening_tight,0.514,0.6296296296296297,0.6149732620320856,0.6446511627906977,0.6608576014675532
-soft_vote_tree_pack,0.511,0.6296296296296297,0.6042780748663101,0.6448959608323135,0.6621187800963081
-random_forest_opening,0.516,0.6296296296296297,0.5935828877005348,0.6448959608323134,0.6563861499656042
-extra_trees_opening_deep,0.514,0.6271604938271605,0.6042780748663101,0.6432558139534884,0.6671634946113276
-soft_vote_opening,0.556,0.6246913580246913,0.6256684491978609,0.638359853121175,0.6513414354505846
-gradient_boost_opening,0.512,0.6246913580246913,0.5828877005347594,0.640734394124847,0.631506535198349
-soft_vote_all_pack,0.503,0.6172839506172839,0.5989304812834224,0.6431089351285189,0.6598257280440265
-catboost_opening,0.524,0.6098765432098765,0.5882352941176471,0.6180660954712363,0.6319651456088053
-hist_gradient_opening,0.517,0.5925925925925926,0.5133689839572193,0.6042839657282741,0.5806007796376977
-logit_opening,0.386,0.582716049382716,0.5454545454545454,0.6403427172582619,0.6141939922036231

backend/models/latest_prediction.csv

@@ -1,2 +0,0 @@
-input_date,first5_start,first5_end,prediction,prob_up,confidence,threshold,model_name,is_overridden
-2026-05-27,2026-05-27 09:15:00,2026-05-27 09:19:00,UP,0.6970470349448922,0.7720470349448922,0.425,blend_extra_trees_tight_logit_overlay,False

backend/models/nifty_1420_tplus1_logistic_model.joblib

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:77fea2ef4be30e6355377c1badadc94e9cb941fb62a31e731fad83b0f171c63e
-size 5321

backend/models/nifty_opening_direction_model.joblib

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:faa463488804279181dc6ee63ca62871d0d5817b0cbb23d34e7374d7628ced98
-size 16234249

backend/models/nifty_tomorrow_direction_model.joblib

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:ce657f6c5de48990fca4621b06796d57a0d71126f159b2adfaa476aaae66ccf6
-size 446

backend/models/refresh_state.json

@@ -1,7 +0,0 @@
-{
-  "phase": "normal",
-  "started_at": "2026-05-25T10:28:13Z",
-  "finished_at": "2026-05-25T10:33:38Z",
-  "session_date": "2026-05-25",
-  "error": null
-}

backend/models/summary.json

@@ -1,29 +0,0 @@
-{
-  "target": "same-day NIFTY 50 close > same-day NIFTY 50 open after first five 1-minute bars",
-  "model_name": "blend_extra_trees_tight_logit_overlay",
-  "threshold": 0.425,
-  "train_rows": 2221,
-  "valid_rows": 405,
-  "test_rows": 187,
-  "train_start": "2015-01-09",
-  "train_end": "2023-12-29",
-  "valid_start": "2024-01-01",
-  "valid_end": "2025-08-14",
-  "test_start": "2025-08-18",
-  "test_end": "2026-05-21",
-  "validation_accuracy": 0.654320987654321,
-  "test_accuracy": 0.6524064171122995,
-  "baseline_test_accuracy": 0.5240641711229946,
-  "validation_auc": 0.6623500611995106,
-  "test_auc": 0.6563861499656042,
-  "validation_log_loss": 0.6563017134616456,
-  "test_log_loss": 0.6607799782446233,
-  "test_brier": 0.2339533732973711,
-  "feature_count": 219,
-  "latest_input_date": "2026-05-21",
-  "latest_first5_start": "2026-05-21 09:15:00",
-  "latest_first5_end": "2026-05-21 09:19:00",
-  "latest_prob_up": 0.4379885393062093,
-  "latest_prediction": "UP",
-  "latest_confidence": 0.5129885393062092
-}

backend/models/tomorrow_latest_prediction.csv

@@ -1,2 +0,0 @@
-input_date,target_date,prediction,prob_up,confidence,threshold,model_name,source_model,validation_accuracy,test_accuracy,artifact_source
-2026-05-27,2026-05-29,UP,0.536599063991701,0.536599063991701,0.534,nifty_tomorrow_direction_model,locked_multiwindow_nifty50_ensemble_v2,0.5780141843971631,0.6736842105263158,C:\Users\jhaji\Downloads\forecasting project\Code\models\nifty_forecaster\outputs

backend/models/tomorrow_summary.json

@@ -1,42 +0,0 @@
-{
-  "symbol": "NIFTY 50",
-  "horizon": "daily",
-  "horizon_bars": 1,
-  "config": {
-    "name": "locked_multiwindow_nifty50_ensemble_v2",
-    "use_intraday": true,
-    "use_external": true,
-    "use_institutional": false,
-    "use_options": true,
-    "use_engineered_macro_flow": false,
-    "blend_mode": "locked_nifty50_multiwindow_v2",
-    "decision_overlay": "bank_body_near_threshold;low_bank_vol_down"
-  },
-  "threshold": 0.534,
-  "validation_accuracy": 0.5780141843971631,
-  "test_accuracy": 0.6736842105263158,
-  "baseline_accuracy": 0.5052631578947369,
-  "n_train": 2221,
-  "n_valid": 282,
-  "n_test": 190,
-  "train_start": "2015-01-09",
-  "train_end": "2023-12-31",
-  "valid_start": "2024-07-01",
-  "valid_end": "2025-08-17",
-  "test_start": "2025-08-18",
-  "test_end": "2026-05-26",
-  "latest_forecast_date": "2026-05-27",
-  "latest_forecast_for": "next trading bar after 2026-05-27",
-  "latest_forecast_prob_up": 0.536599063991701,
-  "latest_forecast_signal": "UP",
-  "feature_count": 204,
-  "validation_prob_std": 0.06800064531350844,
-  "test_prob_std": 0.06311239013827799,
-  "test_prob_min": 0.3874936713840175,
-  "test_prob_max": 0.6430757596651826,
-  "model_name": "nifty_tomorrow_direction_model",
-  "source_model": "locked_multiwindow_nifty50_ensemble_v2",
-  "target": "next trading session NIFTY 50 direction",
-  "artifact_type": "daily_forecaster_outputs",
-  "artifact_source": "C:\\Users\\jhaji\\Downloads\\forecasting project\\Code\\models\\nifty_forecaster\\outputs"
-}

backend/models/tplus1_latest_prediction.csv

@@ -1,2 +0,0 @@
-input_date,target_date,forecast_for,prediction,prob_up,confidence,threshold,model_name,decision_overlay,validation_accuracy,test_accuracy,accuracy_goal
-2026-05-27,2026-05-29,next trading session after 2026-05-27,UP,0.47376564177714775,0.5262343582228522,0.578,logistic_regression_l1_C0.35_balanced,prev_target_mean10_le_0.4_up;m02_range_1m_ge_0.000479116_up,0.66,0.6368421052631579,0.63

backend/models/tplus1_summary.json

@@ -1,37 +0,0 @@
-{
-  "target": "T+1 NIFTY 50 close greater than T 14:20 close",
-  "exchange_instrument": "NSE NIFTY 50 index",
-  "window_start": "14:00",
-  "window_end": "14:20",
-  "lookback_days_requested": 1000,
-  "supervised_rows": 1000,
-  "train_rows": 660,
-  "valid_rows": 150,
-  "test_rows": 190,
-  "train_start": "2022-04-28",
-  "train_end": "2024-12-31",
-  "valid_start": "2025-01-01",
-  "valid_end": "2025-08-06",
-  "test_start": "2025-08-07",
-  "test_end": "2026-05-20",
-  "model_name": "logistic_regression_l1_C0.35_balanced",
-  "threshold": 0.578,
-  "validation_accuracy": 0.66,
-  "test_accuracy": 0.6368421052631579,
-  "baseline_test_accuracy": 0.5052631578947369,
-  "validation_auc": 0.5085348506401137,
-  "test_auc": 0.54864804964539,
-  "test_log_loss": 0.715200083567372,
-  "test_brier": 0.2580990249203714,
-  "accuracy_goal": 0.63,
-  "accuracy_goal_met_on_test": true,
-  "feature_count": 40,
-  "latest_input_date": "2026-05-21",
-  "latest_window_rows": 21,
-  "latest_forecast_for": "next trading session after 2026-05-21",
-  "latest_prob_up": 0.620604654276822,
-  "latest_prediction": "UP",
-  "latest_confidence": 0.620604654276822,
-  "decision_overlay": "prev_target_mean10_le_0.4_up;m02_range_1m_ge_0.000479116_up",
-  "top_features": 40
-}

backend/models/yahoo_history_cache.sqlite3

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:534853b3b5984c2c4128ed073412f090ce96d6cb7a9da4aaa98a20482a2a4bff
-size 155648

backend/nifty_backend/__init__.py

@@ -1,2 +0,0 @@
-"""Backend runtime for the NIFTY 50 opening-direction forecaster."""
-

backend/nifty_backend/__pycache__/runtime.cpython-311.pyc

@@ -1,3 +0,0 @@
-version https://git-lfs.github.com/spec/v1
-oid sha256:89f2a00f920ffc636fbdd9b8b9b0fb6c6cbe955a5de208b7879183571d861e84
-size 112925

backend/nifty_backend/runtime.py

@@ -1,1632 +0,0 @@
-from __future__ import annotations
-
-import json
-import copy
-import os
-import sys
-import threading
-from dataclasses import dataclass
-from datetime import date, datetime, time, timedelta
-from functools import lru_cache
-from pathlib import Path
-from typing import Any
-from zoneinfo import ZoneInfo
-
-import joblib
-import numpy as np
-import pandas as pd
-from nifty_backend.yahoo_history_client import YahooHistoryClient
-
-try:
-    import pandas_market_calendars as mcal
-except ImportError:  # pragma: no cover - production dependency, local fallback below.
-    mcal = None
-
-
-IST = ZoneInfo("Asia/Kolkata")
-YAHOO_NIFTY_SYMBOL = "^NSEI"
-MARKET_CLOSE = time(15, 30)
-FIRST5_READY = time(9, 20)
-CLOSE_REFRESH_READY = time(15, 45)
-TPLUS1_READY = time(14, 30)
-STALE_CHECK_INTERVAL_SECONDS = 5
-BACKEND_ROOT = Path(__file__).resolve().parents[1]
-DATA_DIR = BACKEND_ROOT / "data"
-MODEL_DIR = BACKEND_ROOT / "models"
-YAHOO_CACHE_PATH = MODEL_DIR / "yahoo_history_cache.sqlite3"
-OPENING_DATASET_PATH = DATA_DIR / "opening_direction_training_dataset.parquet"
-NIFTY_1M_PATH = DATA_DIR / "nifty50_1m.parquet"
-NIFTY_1D_PATH = DATA_DIR / "nifty50_1d.parquet"
-MODEL_PATH = MODEL_DIR / "nifty_opening_direction_model.joblib"
-LATEST_PATH = MODEL_DIR / "latest_prediction.csv"
-TEST_PREDICTIONS_PATH = DATA_DIR / "test_predictions.parquet"
-TOMORROW_MODEL_PATH = MODEL_DIR / "nifty_tomorrow_direction_model.joblib"
-TOMORROW_LATEST_PATH = MODEL_DIR / "tomorrow_latest_prediction.csv"
-TOMORROW_SUMMARY_PATH = MODEL_DIR / "tomorrow_summary.json"
-TOMORROW_TEST_PREDICTIONS_PATH = DATA_DIR / "tomorrow_test_predictions.parquet"
-FORECASTING_PROJECT_ROOT = Path(
-    os.environ.get(
-        "FORECASTING_PROJECT_ROOT",
-        str(BACKEND_ROOT.parent.parent / "forecasting project"),
-    )
-)
-DAILY_FORECASTER_OUTPUT_DIR = FORECASTING_PROJECT_ROOT / "Code" / "models" / "nifty_forecaster" / "outputs"
-DAILY_FORECASTER_SUMMARY_PATH = DAILY_FORECASTER_OUTPUT_DIR / "forecaster_summary.json"
-DAILY_FORECASTER_LATEST_PATH = DAILY_FORECASTER_OUTPUT_DIR / "forecaster_latest.csv"
-DAILY_FORECASTER_PREDICTIONS_PATH = DAILY_FORECASTER_OUTPUT_DIR / "forecaster_test_predictions.csv"
-TPLUS1_MODEL_PATH = MODEL_DIR / "nifty_1420_tplus1_logistic_model.joblib"
-TPLUS1_LATEST_PATH = MODEL_DIR / "tplus1_latest_prediction.csv"
-TPLUS1_SUMMARY_PATH = MODEL_DIR / "tplus1_summary.json"
-TPLUS1_TEST_PREDICTIONS_PATH = DATA_DIR / "tplus1_test_predictions.parquet"
-REFRESH_STATE_PATH = MODEL_DIR / "refresh_state.json"
-REFRESH_WAITING = "waiting_second_payload"
-REFRESH_REFRESHING = "refreshing"
-REFRESH_READY = "ready"
-REFRESH_FAILED = "failed"
-REFRESH_NORMAL = "normal"
-LIVE_ACCURACY_PATH = MODEL_DIR / "live_accuracy.json"
-
-DECISION_OVERLAYS = [
-    {
-        "name": "fifth_minute_momentum_flip",
-        "feature": "m5_ret_1m",
-        "op": ">=",
-        "value": 0.0005085411885759201,
-    },
-    {
-        "name": "vix_stretch_flip",
-        "feature": "india_vix_close_vs_sma_20",
-        "op": ">=",
-        "value": 0.24641908937959742,
-    },
-]
-
-_dashboard_payload_lock = threading.Lock()
-_stale_refresh_lock = threading.Lock()
-
-
-def utc_now_iso() -> str:
-    return datetime.utcnow().replace(microsecond=0).isoformat() + "Z"
-
-
-def clear_dashboard_payload_cache() -> None:
-    _dashboard_payload_cached.cache_clear()
-
-
-def save_refresh_state(phase: str, *, session_date: date | None = None, error: str | None = None) -> dict[str, Any]:
-    previous = load_refresh_state()
-    state = {
-        "phase": phase,
-        "started_at": previous.get("started_at"),
-        "finished_at": previous.get("finished_at"),
-        "session_date": session_date.isoformat() if session_date else previous.get("session_date"),
-        "error": error,
-    }
-    if phase in {REFRESH_WAITING, REFRESH_REFRESHING} and not state["started_at"]:
-        state["started_at"] = utc_now_iso()
-    if phase in {REFRESH_READY, REFRESH_FAILED, REFRESH_NORMAL}:
-        state["finished_at"] = utc_now_iso()
-    REFRESH_STATE_PATH.write_text(json.dumps(state, indent=2), encoding="utf-8")
-    return state
-
-
-def load_refresh_state() -> dict[str, Any]:
-    if not REFRESH_STATE_PATH.exists():
-        return {
-            "phase": REFRESH_NORMAL,
-            "started_at": None,
-            "finished_at": None,
-            "session_date": None,
-            "error": None,
-        }
-    try:
-        return json.loads(REFRESH_STATE_PATH.read_text(encoding="utf-8"))
-    except Exception:
-        return {
-            "phase": REFRESH_FAILED,
-            "started_at": None,
-            "finished_at": None,
-            "session_date": None,
-            "error": "refresh_state.json could not be read",
-        }
-
-
-@lru_cache(maxsize=1)
-def _nse_calendar():
-    if mcal is None:
-        return None
-    for name in ("XNSE", "NSE", "BSE"):
-        try:
-            return mcal.get_calendar(name)
-        except Exception:
-            continue
-    return None
-
-
-@lru_cache(maxsize=64)
-def trading_schedule(start: date, end: date) -> pd.DataFrame:
-    calendar = _nse_calendar()
-    if calendar is None:
-        days = pd.date_range(start=start, end=end, freq="B")
-        return pd.DataFrame(index=days)
-    return calendar.schedule(start_date=start, end_date=end)
-
-
-def is_trading_day(day: date) -> bool:
-    schedule = trading_schedule(day, day)
-    return not schedule.empty
-
-
-def next_trading_day(start: date) -> date:
-    end = start + timedelta(days=14)
-    schedule = trading_schedule(start, end)
-    if schedule.empty:
-        day = start
-        while not is_trading_day(day):
-            day += timedelta(days=1)
-        return day
-    return pd.Timestamp(schedule.index[0]).date()
-
-
-def previous_trading_day(start: date) -> date:
-    begin = start - timedelta(days=14)
-    schedule = trading_schedule(begin, start)
-    if schedule.empty:
-        day = start
-        while not is_trading_day(day):
-            day -= timedelta(days=1)
-        return day
-    return pd.Timestamp(schedule.index[-1]).date()
-
-
-class ProbabilityBlend:
-    def __init__(self, models: list[Any], weights: np.ndarray):
-        self.models = models
-        self.weights = np.asarray(weights, dtype="float64")
-        self.weights = self.weights / self.weights.sum()
-
-    def predict_proba(self, x: pd.DataFrame) -> np.ndarray:
-        probs = np.column_stack([predict_proba_up(model, x) for model in self.models])
-        prob_up = probs @ self.weights
-        return np.column_stack([1.0 - prob_up, prob_up])
-
-
-@dataclass(frozen=True)
-class Prediction:
-    input_date: str
-    first5_start: str
-    first5_end: str
-    prediction: str
-    prob_up: float
-    confidence: float
-    threshold: float
-    model_name: str
-    is_overridden: bool = False
-
-    def to_dict(self) -> dict[str, Any]:
-        return {
-            "input_date": self.input_date,
-            "first5_start": self.first5_start,
-            "first5_end": self.first5_end,
-            "prediction": self.prediction,
-            "prob_up": self.prob_up,
-            "confidence": self.confidence,
-            "threshold": self.threshold,
-            "model_name": self.model_name,
-            "is_overridden": getattr(self, "is_overridden", False),
-        }
-
-
-def predict_proba_up(model: Any, x: pd.DataFrame) -> np.ndarray:
-    return np.asarray(model.predict_proba(x)[:, 1], dtype="float64")
-
-
-def safe_div(numer: pd.Series | np.ndarray, denom: pd.Series | np.ndarray) -> pd.Series:
-    n = pd.Series(numer, copy=False)
-    d = pd.Series(denom, copy=False)
-    out = pd.Series(np.nan, index=n.index, dtype="float64")
-    mask = d.notna() & np.isfinite(d.to_numpy(dtype="float64")) & (d != 0)
-    out.loc[mask] = n.loc[mask].to_numpy(dtype="float64") / d.loc[mask].to_numpy(dtype="float64")
-    return out
-
-
-def load_model() -> dict[str, Any]:
-    # Existing artifact was trained as a script, so its custom blend class
-    # resolves through __main__ when unpickled.
-    sys.modules["__main__"].ProbabilityBlend = ProbabilityBlend
-    sys.modules["__main__"].predict_proba_up = predict_proba_up
-    payload = joblib.load(MODEL_PATH)
-    payload.setdefault("decision_overlays", DECISION_OVERLAYS)
-    payload.setdefault("model_name", "nifty_opening_direction_model")
-    return payload
-
-
-def overlay_mask(frame: pd.DataFrame, overlay: dict[str, object]) -> np.ndarray:
-    feature = str(overlay["feature"])
-    if feature not in frame.columns:
-        return np.zeros(len(frame), dtype=bool)
-    series = pd.to_numeric(frame[feature], errors="coerce")
-    value = float(overlay["value"])
-    if overlay["op"] == ">=":
-        return (series >= value).fillna(False).to_numpy(dtype=bool)
-    if overlay["op"] == "<=":
-        return (series <= value).fillna(False).to_numpy(dtype=bool)
-    raise ValueError(f"Unsupported overlay op: {overlay['op']}")
-
-
-def apply_decision_overlays(pred: np.ndarray, frame: pd.DataFrame, overlays: list[dict[str, object]]) -> np.ndarray:
-    adjusted = np.asarray(pred, dtype="int64").copy()
-    for overlay in overlays:
-        mask = overlay_mask(frame, overlay)
-        adjusted[mask] = 1 - adjusted[mask]
-    return adjusted
-
-
-def directional_confidence(prob_up: np.ndarray, pred: np.ndarray, threshold: float) -> np.ndarray:
-    prob_up = np.asarray(prob_up, dtype="float64")
-    pred = np.asarray(pred, dtype="int64")
-    base_side_prob = np.where(pred == 1, prob_up, 1.0 - prob_up)
-    threshold_distance = np.abs(prob_up - float(threshold))
-    return np.clip(0.50 + threshold_distance, base_side_prob, 0.99)
-
-
-def read_training_dataset() -> pd.DataFrame:
-    df = pd.read_parquet(OPENING_DATASET_PATH)
-    for col in ("date", "first5_start", "first5_end"):
-        if col in df.columns:
-            df[col] = pd.to_datetime(df[col], errors="coerce")
-    return df.sort_values("date").reset_index(drop=True)
-
-
-def normalize_yahoo_frame(df: pd.DataFrame) -> pd.DataFrame:
-    if df.empty:
-        return pd.DataFrame(columns=["date", "open", "high", "low", "close", "volume"])
-    if isinstance(df.columns, pd.MultiIndex):
-        df.columns = [str(c[0]).lower() for c in df.columns]
-    else:
-        df.columns = [str(c).lower().replace(" ", "_") for c in df.columns]
-    df = df.reset_index()
-    date_col = next((c for c in df.columns if c.lower() in {"datetime", "date"}), df.columns[0])
-    df["date"] = pd.to_datetime(df[date_col], errors="coerce")
-    if df["date"].dt.tz is None:
-        df["date"] = df["date"].dt.tz_localize("UTC").dt.tz_convert(IST)
-    else:
-        df["date"] = df["date"].dt.tz_convert(IST)
-    rename = {
-        "open": "open",
-        "high": "high",
-        "low": "low",
-        "close": "close",
-        "adj_close": "close",
-        "volume": "volume",
-    }
-    out = pd.DataFrame({"date": df["date"].dt.tz_localize(None)})
-    for src, dst in rename.items():
-        if src in df.columns and dst not in out.columns:
-            out[dst] = pd.to_numeric(df[src], errors="coerce")
-    return out.dropna(subset=["date", "open", "high", "low", "close"]).sort_values("date")
-
-
-@lru_cache(maxsize=1)
-def yahoo_history_client() -> YahooHistoryClient:
-    return YahooHistoryClient(cache_path=YAHOO_CACHE_PATH)
-
-
-def period_start(period: str, *, end: datetime) -> datetime:
-    text = str(period).strip().lower()
-    units = {
-        "d": "days",
-        "wk": "weeks",
-        "mo": "months",
-        "y": "years",
-    }
-    for suffix, unit in units.items():
-        if text.endswith(suffix):
-            raw_value = text[: -len(suffix)]
-            if not raw_value.isdigit():
-                break
-            value = int(raw_value)
-            if unit == "days":
-                return end - timedelta(days=value)
-            if unit == "weeks":
-                return end - timedelta(weeks=value)
-            if unit == "months":
-                return end - timedelta(days=value * 31)
-            if unit == "years":
-                return end - timedelta(days=value * 366)
-    raise ValueError(f"Unsupported Yahoo period: {period!r}")
-
-
-def yahoo_history_to_ohlcv(frame: pd.DataFrame, *, daily: bool) -> pd.DataFrame:
-    if frame.empty:
-        return pd.DataFrame(columns=["date", "open", "high", "low", "close", "volume"])
-    out = frame.rename(columns={"timestamp": "date"}).copy()
-    out["date"] = pd.to_datetime(out["date"], errors="coerce")
-    if daily:
-        out["date"] = out["date"].dt.normalize()
-    for column in ("open", "high", "low", "close", "volume"):
-        out[column] = pd.to_numeric(out[column], errors="coerce")
-    return (
-        out[["date", "open", "high", "low", "close", "volume"]]
-        .dropna(subset=["date", "open", "high", "low", "close"])
-        .drop_duplicates("date", keep="last")
-        .sort_values("date")
-        .reset_index(drop=True)
-    )
-
-
-def fetch_yahoo_minutes(period: str = "5d") -> pd.DataFrame:
-    end = datetime.now(IST).replace(tzinfo=None) + timedelta(minutes=5)
-    start = period_start(period, end=end)
-    raw = yahoo_history_client().fetch_history(
-        YAHOO_NIFTY_SYMBOL,
-        interval="1m",
-        start=start,
-        end=end,
-        include_prepost=False,
-    )
-    return yahoo_history_to_ohlcv(raw, daily=False)
-
-
-def fetch_yahoo_daily(period: str = "1mo") -> pd.DataFrame:
-    end = datetime.now(IST).replace(tzinfo=None) + timedelta(days=1)
-    start = period_start(period, end=end)
-    raw = yahoo_history_client().fetch_history(
-        YAHOO_NIFTY_SYMBOL,
-        interval="1d",
-        start=start,
-        end=end,
-        include_prepost=False,
-    )
-    return yahoo_history_to_ohlcv(raw, daily=True)
-
-
-def append_parquet_rows(path: Path, new_rows: pd.DataFrame, subset: list[str]) -> pd.DataFrame:
-    if new_rows.empty:
-        if path.exists():
-            return pd.read_parquet(path)
-        raise RuntimeError(f"No rows returned for {path.name}; leaving parquet unchanged.")
-    if path.exists():
-        existing = pd.read_parquet(path)
-        combined = pd.concat([existing, new_rows], ignore_index=True)
-    else:
-        combined = new_rows.copy()
-    combined = combined.drop_duplicates(subset=subset, keep="last").sort_values(subset).reset_index(drop=True)
-    combined.to_parquet(path, index=False, compression="zstd")
-    return combined
-
-
-def latest_parquet_date(path: Path) -> date | None:
-    if not path.exists():
-        return None
-    df = pd.read_parquet(path, columns=["date"])
-    if df.empty:
-        return None
-    latest = pd.to_datetime(df["date"], errors="coerce").max()
-    if pd.isna(latest):
-        return None
-    return latest.date()
-
-
-def latest_opening_outcome_date() -> date | None:
-    if not OPENING_DATASET_PATH.exists():
-        return None
-    cols = ["date"]
-    if "target" in pd.read_parquet(OPENING_DATASET_PATH).columns:
-        cols.append("target")
-    df = pd.read_parquet(OPENING_DATASET_PATH, columns=cols)
-    if df.empty or "target" not in df.columns:
-        return None
-    df = df[df["target"].notna()]
-    if df.empty:
-        return None
-    latest = pd.to_datetime(df["date"], errors="coerce").max()
-    if pd.isna(latest):
-        return None
-    return latest.date()
-
-
-def first5_features_from_minutes(minutes: pd.DataFrame, session_date: date | None = None) -> pd.DataFrame:
-    if minutes.empty:
-        raise RuntimeError("Yahoo returned no minute bars.")
-    bars = minutes.copy()
-    bars["dt"] = pd.to_datetime(bars["date"], errors="coerce")
-    bars["session_date"] = bars["dt"].dt.normalize()
-    if session_date is None:
-        session_ts = bars["session_date"].max()
-    else:
-        session_ts = pd.Timestamp(session_date).normalize()
-    day = bars[bars["session_date"] == session_ts].sort_values("dt").copy()
-    start_dt = pd.Timestamp.combine(session_ts.date(), time(9, 15))
-    end_dt = pd.Timestamp.combine(session_ts.date(), time(9, 19))
-    first5 = day[(day["dt"] >= start_dt) & (day["dt"] <= end_dt)].head(5).copy()
-    if len(first5) < 5:
-        raise RuntimeError(f"Need 5 opening bars for {session_ts.date()}, got {len(first5)}.")
-    first5["minute_index"] = np.arange(len(first5))
-    first5["ret_1m"] = first5["close"].pct_change(fill_method=None)
-    first5["range_pct_1m"] = safe_div(first5["high"] - first5["low"], first5["open"])
-    first5["body_pct_1m"] = safe_div(first5["close"] - first5["open"], first5["open"])
-    row = {
-        "date": session_ts,
-        "first5_start": first5["dt"].iloc[0],
-        "first5_end": first5["dt"].iloc[-1],
-        "first5_open": first5["open"].iloc[0],
-        "first5_high": first5["high"].max(),
-        "first5_low": first5["low"].min(),
-        "first5_close": first5["close"].iloc[-1],
-        "first5_volume": first5["volume"].sum() if "volume" in first5 else 0.0,
-        "first5_bars": len(first5),
-        "first5_last_1m_ret": first5["ret_1m"].iloc[-1],
-        "first5_ret_std": first5["ret_1m"].std(),
-    }
-    row["first5_return"] = (row["first5_close"] - row["first5_open"]) / row["first5_open"]
-    row["first5_range_pct"] = (row["first5_high"] - row["first5_low"]) / row["first5_open"]
-    first5_range = row["first5_high"] - row["first5_low"]
-    row["first5_body_to_range"] = (row["first5_close"] - row["first5_open"]) / first5_range if first5_range else np.nan
-    row["first5_close_location"] = (row["first5_close"] - row["first5_low"]) / first5_range if first5_range else np.nan
-    for idx, (_, candle) in enumerate(first5.iterrows(), start=1):
-        for field in ("open", "high", "low", "close", "ret_1m", "range_pct_1m", "body_pct_1m"):
-            row[f"m{idx}_{field}"] = candle[field]
-        row[f"m{idx}_close_vs_first5_open"] = (candle["close"] - row["first5_open"]) / row["first5_open"]
-        row[f"m{idx}_range_share"] = (candle["high"] - candle["low"]) / first5_range if first5_range else np.nan
-    row["first5_return_accel"] = row["m5_ret_1m"] - row["m2_ret_1m"]
-    row["first5_last2_return"] = (row["m5_close"] - row["m4_open"]) / row["m4_open"]
-    row["first5_first2_return"] = (row["m2_close"] - row["m1_open"]) / row["m1_open"]
-    row["first5_reversal"] = np.sign(row["first5_first2_return"]) * -np.sign(row["first5_last2_return"])
-    row["dow"] = session_ts.dayofweek
-    row["dom"] = session_ts.day
-    row["month"] = session_ts.month
-    return pd.DataFrame([row])
-
-
-def build_model_row(first5_row: pd.DataFrame) -> pd.DataFrame:
-    dataset = read_training_dataset()
-    latest_context = dataset.iloc[[-1]].copy()
-    output = latest_context.copy()
-    for col in first5_row.columns:
-        output[col] = first5_row[col].iloc[0]
-    if {"first5_open", "nifty_close"}.issubset(output.columns):
-        output["first5_gap_from_prev_close"] = (output["first5_open"] - output["nifty_close"]) / output["nifty_close"]
-        output["first5_close_vs_prev_close"] = (output["first5_close"] - output["nifty_close"]) / output["nifty_close"]
-    if {"first5_range_pct", "nifty_range_pct"}.issubset(output.columns):
-        output["first5_range_vs_prev_range"] = output["first5_range_pct"] / output["nifty_range_pct"]
-    if {"first5_return", "nifty_ret_1"}.issubset(output.columns):
-        output["first5_return_x_prev_ret"] = output["first5_return"] * output["nifty_ret_1"]
-        output["gap_x_prev_ret"] = output["first5_gap_from_prev_close"] * output["nifty_ret_1"]
-    if {"first5_return", "banknifty_ret_1"}.issubset(output.columns):
-        output["first5_return_x_bank_ret_1"] = output["first5_return"] * output["banknifty_ret_1"]
-    if {"first5_range_pct", "india_vix_ret_1"}.issubset(output.columns):
-        output["first5_range_x_vix_ret_1"] = output["first5_range_pct"] * output["india_vix_ret_1"]
-    output["target"] = np.nan
-    output["day_return"] = np.nan
-    return output
-
-
-def predict_row(row: pd.DataFrame) -> Prediction:
-    payload = load_model()
-    model = payload["model"]
-    features = payload["features"]
-    threshold = float(payload["threshold"])
-    missing = [c for c in features if c not in row.columns]
-    if missing:
-        raise RuntimeError(f"Feature row is missing {len(missing)} features; first missing: {missing[:5]}")
-    prob_up = predict_proba_up(model, row[features])
-    raw_pred = (prob_up >= threshold).astype("int64")
-    pred = apply_decision_overlays(raw_pred, row, payload.get("decision_overlays", DECISION_OVERLAYS))
-    is_overridden = bool(raw_pred[0] != pred[0])
-    confidence = directional_confidence(prob_up, pred, threshold)
-    prediction = Prediction(
-        input_date=pd.to_datetime(row["date"].iloc[0]).date().isoformat(),
-        first5_start=str(pd.to_datetime(row["first5_start"].iloc[0])),
-        first5_end=str(pd.to_datetime(row["first5_end"].iloc[0])),
-        prediction="UP" if int(pred[0]) == 1 else "DOWN",
-        prob_up=float(prob_up[0]),
-        confidence=float(confidence[0]),
-        threshold=threshold,
-        model_name=str(payload.get("model_name", "nifty_opening_direction_model")),
-        is_overridden=is_overridden,
-    )
-    pd.DataFrame([prediction.to_dict()]).to_csv(LATEST_PATH, index=False)
-    return prediction
-
-
-def _file_cache_key(path: Path) -> tuple[str, int | None, int | None]:
-    try:
-        stat = path.stat()
-    except FileNotFoundError:
-        return (str(path), None, None)
-    return (str(path), stat.st_mtime_ns, stat.st_size)
-
-
-@lru_cache(maxsize=16)
-def _latest_saved_prediction_cached(latest_key: tuple[str, int | None, int | None], summary_key: tuple[str, int | None, int | None]) -> dict[str, Any]:
-    latest_path = Path(latest_key[0])
-    if latest_path.exists():
-        return pd.read_csv(latest_path).iloc[-1].to_dict()
-    summary_path = Path(summary_key[0])
-    if summary_path.exists():
-        return json.loads(summary_path.read_text(encoding="utf-8"))
-    raise FileNotFoundError("No latest prediction is available yet.")
-
-
-def latest_saved_prediction() -> dict[str, Any]:
-    return dict(_latest_saved_prediction_cached(_file_cache_key(LATEST_PATH), _file_cache_key(MODEL_DIR / "summary.json")))
-
-
-def _latest_saved_prediction_uncached() -> dict[str, Any]:
-    if LATEST_PATH.exists():
-        return pd.read_csv(LATEST_PATH).iloc[-1].to_dict()
-    summary_path = MODEL_DIR / "summary.json"
-    if summary_path.exists():
-        return json.loads(summary_path.read_text(encoding="utf-8"))
-    raise FileNotFoundError("No latest prediction is available yet.")
-
-
-def _read_daily_forecaster_summary() -> dict[str, Any] | None:
-    if not DAILY_FORECASTER_SUMMARY_PATH.exists():
-        return None
-    raw = json.loads(DAILY_FORECASTER_SUMMARY_PATH.read_text(encoding="utf-8"))
-    if isinstance(raw, list):
-        matches = [row for row in raw if row.get("symbol") == "NIFTY 50"]
-        summary = dict(matches[0] if matches else raw[0])
-    elif isinstance(raw, dict):
-        summary = dict(raw)
-    else:
-        return None
-    config = summary.get("config") if isinstance(summary.get("config"), dict) else {}
-    summary.setdefault("symbol", "NIFTY 50")
-    summary.setdefault("horizon", "daily")
-    summary.setdefault("horizon_bars", 1)
-    summary["model_name"] = "nifty_tomorrow_direction_model"
-    summary["source_model"] = str(config.get("name") or summary.get("source_model") or "locked_multiwindow_nifty50_ensemble")
-    summary["target"] = "next trading session NIFTY 50 direction"
-    summary["artifact_type"] = "daily_forecaster_outputs"
-    summary["artifact_source"] = str(DAILY_FORECASTER_OUTPUT_DIR)
-    return summary
-
-
-def _read_daily_forecaster_latest(summary: dict[str, Any]) -> dict[str, Any] | None:
-    if not DAILY_FORECASTER_LATEST_PATH.exists():
-        return None
-    latest = pd.read_csv(DAILY_FORECASTER_LATEST_PATH)
-    if latest.empty:
-        return None
-    if "symbol" in latest.columns:
-        filtered = latest[latest["symbol"].astype(str) == "NIFTY 50"]
-        if not filtered.empty:
-            latest = filtered
-    row = {k: (None if pd.isna(v) else v) for k, v in latest.iloc[-1].to_dict().items()}
-    input_date = row.get("latest_forecast_date") or row.get("input_date")
-    target_date = row.get("target_date")
-    if not target_date and input_date:
-        try:
-            target_date = next_trading_day(date.fromisoformat(str(input_date)[:10]) + timedelta(days=1)).isoformat()
-        except Exception:
-            target_date = None
-    prob_up = row.get("latest_forecast_prob_up", row.get("prob_up"))
-    prediction = row.get("latest_forecast_signal", row.get("prediction"))
-    threshold = row.get("threshold", summary.get("threshold"))
-    confidence = row.get("confidence")
-    if confidence is None and prob_up is not None:
-        try:
-            confidence = float(max(float(prob_up), 1.0 - float(prob_up)))
-        except Exception:
-            confidence = None
-    return {
-        "input_date": input_date,
-        "target_date": target_date,
-        "prediction": prediction,
-        "prob_up": prob_up,
-        "confidence": confidence,
-        "threshold": threshold,
-        "model_name": "nifty_tomorrow_direction_model",
-        "source_model": summary.get("source_model", "locked_multiwindow_nifty50_ensemble"),
-        "validation_accuracy": summary.get("validation_accuracy"),
-        "test_accuracy": summary.get("test_accuracy"),
-        "artifact_source": str(DAILY_FORECASTER_OUTPUT_DIR),
-    }
-
-
-def sync_daily_forecaster_outputs() -> dict[str, Any] | None:
-    summary = _read_daily_forecaster_summary()
-    if summary is None:
-        return None
-    latest = _read_daily_forecaster_latest(summary)
-    TOMORROW_SUMMARY_PATH.write_text(json.dumps(summary, indent=2), encoding="utf-8")
-    if latest is not None:
-        pd.DataFrame([latest]).to_csv(TOMORROW_LATEST_PATH, index=False)
-    if DAILY_FORECASTER_PREDICTIONS_PATH.exists():
-        predictions = pd.read_csv(DAILY_FORECASTER_PREDICTIONS_PATH)
-        if "symbol" in predictions.columns:
-            predictions = predictions[predictions["symbol"].astype(str) == "NIFTY 50"].copy()
-        if not predictions.empty:
-            if "pred" in predictions.columns and "prediction" not in predictions.columns:
-                predictions["prediction"] = np.where(pd.to_numeric(predictions["pred"], errors="coerce") == 1, "UP", "DOWN")
-            if "correct" not in predictions.columns and {"target", "pred"}.issubset(predictions.columns):
-                predictions["correct"] = (
-                    pd.to_numeric(predictions["target"], errors="coerce")
-                    == pd.to_numeric(predictions["pred"], errors="coerce")
-                )
-            predictions.to_parquet(TOMORROW_TEST_PREDICTIONS_PATH, index=False)
-    artifact = {
-        "artifact_type": "daily_forecaster_outputs",
-        "model_name": "nifty_tomorrow_direction_model",
-        "source_model": summary.get("source_model", "locked_multiwindow_nifty50_ensemble"),
-        "threshold": float(summary.get("threshold", 0.54)),
-        "validation_accuracy": summary.get("validation_accuracy"),
-        "test_accuracy": summary.get("test_accuracy"),
-        "validation_prob_std": summary.get("validation_prob_std"),
-        "test_prob_std": summary.get("test_prob_std"),
-        "test_prob_min": summary.get("test_prob_min"),
-        "test_prob_max": summary.get("test_prob_max"),
-        "artifact_source": str(DAILY_FORECASTER_OUTPUT_DIR),
-    }
-    joblib.dump(artifact, TOMORROW_MODEL_PATH)
-    return latest or summary
-
-
-def load_tomorrow_model_artifact() -> dict[str, Any]:
-    synced = sync_daily_forecaster_outputs()
-    if synced is not None and TOMORROW_MODEL_PATH.exists():
-        return joblib.load(TOMORROW_MODEL_PATH)
-    if TOMORROW_MODEL_PATH.exists():
-        return joblib.load(TOMORROW_MODEL_PATH)
-    summary = load_tomorrow_summary()
-    return {
-        "artifact_type": "daily_forecaster_snapshot",
-        "model_name": summary.get("model_name", "nifty_tomorrow_direction_model"),
-        "source_model": summary.get("source_model", "tuned_daily_forest_single"),
-        "threshold": float(summary.get("threshold", 0.543)),
-    }
-
-
-def load_tomorrow_summary() -> dict[str, Any]:
-    synced = sync_daily_forecaster_outputs()
-    if synced is not None and TOMORROW_SUMMARY_PATH.exists():
-        return json.loads(TOMORROW_SUMMARY_PATH.read_text(encoding="utf-8"))
-    if TOMORROW_SUMMARY_PATH.exists():
-        return json.loads(TOMORROW_SUMMARY_PATH.read_text(encoding="utf-8"))
-    return {
-        "model_name": "nifty_tomorrow_direction_model",
-        "source_model": "locked_multiwindow_nifty50_ensemble",
-        "target": "next trading session NIFTY 50 direction",
-        "threshold": 0.54,
-        "validation_accuracy": 0.5673758865248227,
-        "test_accuracy": 0.6451612903225806,
-        "baseline_accuracy": 0.5053763440860215,
-        "n_test": 186,
-        "feature_count": 204,
-    }
-
-
-def latest_tomorrow_prediction() -> dict[str, Any]:
-    sync_daily_forecaster_outputs()
-    latest_daily = latest_parquet_date(NIFTY_1D_PATH)
-    expected_daily = expected_completed_daily_date()
-    valid_daily = min(latest_daily, expected_daily) if latest_daily and expected_daily else (expected_daily or latest_daily)
-
-    if TOMORROW_LATEST_PATH.exists():
-        row = pd.read_csv(TOMORROW_LATEST_PATH).iloc[-1].to_dict()
-        cleaned = {k: (None if pd.isna(v) else v) for k, v in row.items()}
-        try:
-            input_day = date.fromisoformat(str(cleaned.get("input_date"))[:10])
-        except Exception:
-            input_day = None
-        if valid_daily is not None and (input_day is None or input_day < valid_daily):
-            try:
-                refreshed = refresh_tomorrow_prediction(session_date=valid_daily)
-                try:
-                    refreshed_day = date.fromisoformat(str(refreshed.get("input_date"))[:10])
-                except Exception:
-                    refreshed_day = None
-                if refreshed_day is not None and refreshed_day >= valid_daily:
-                    return refreshed
-            except Exception:
-                pass
-        return cleaned
-    summary = load_tomorrow_summary()
-    try:
-        summary_input_day = date.fromisoformat(str(summary.get("latest_forecast_date"))[:10])
-    except Exception:
-        summary_input_day = None
-    if valid_daily is not None and (summary_input_day is None or summary_input_day < valid_daily):
-        try:
-            return refresh_tomorrow_prediction(session_date=valid_daily)
-        except Exception:
-            pass
-    return {
-        "input_date": summary.get("latest_forecast_date"),
-        "target_date": None,
-        "prediction": summary.get("latest_forecast_signal"),
-        "prob_up": summary.get("latest_forecast_prob_up"),
-        "confidence": None,
-        "threshold": summary.get("threshold"),
-        "model_name": summary.get("model_name", "nifty_tomorrow_direction_model"),
-        "source_model": summary.get("source_model", "tuned_daily_forest_single"),
-        "validation_accuracy": summary.get("validation_accuracy"),
-        "test_accuracy": summary.get("test_accuracy"),
-    }
-
-
-def load_tplus1_summary() -> dict[str, Any]:
-    if TPLUS1_SUMMARY_PATH.exists():
-        return json.loads(TPLUS1_SUMMARY_PATH.read_text(encoding="utf-8"))
-    return {
-        "model_name": "logistic_regression_l1_C0.35_balanced",
-        "target": "T+1 NIFTY 50 close greater than T 14:20 close",
-        "window_start": "14:00",
-        "window_end": "14:20",
-        "threshold": 0.578,
-        "validation_accuracy": 0.66,
-        "test_accuracy": 0.6368421052631579,
-        "baseline_test_accuracy": 0.5052631578947369,
-        "test_rows": 190,
-        "feature_count": 40,
-    }
-
-
-def latest_tplus1_prediction() -> dict[str, Any]:
-    if TPLUS1_LATEST_PATH.exists():
-        row = pd.read_csv(TPLUS1_LATEST_PATH).iloc[-1].to_dict()
-        return {k: (None if pd.isna(v) else v) for k, v in row.items()}
-    summary = load_tplus1_summary()
-    return {
-        "input_date": summary.get("latest_input_date"),
-        "target_date": None,
-        "forecast_for": summary.get("latest_forecast_for"),
-        "prediction": summary.get("latest_prediction"),
-        "prob_up": summary.get("latest_prob_up"),
-        "confidence": summary.get("latest_confidence"),
-        "threshold": summary.get("threshold"),
-        "model_name": summary.get("model_name", "logistic_regression_l1_C0.35_balanced"),
-        "validation_accuracy": summary.get("validation_accuracy"),
-        "test_accuracy": summary.get("test_accuracy"),
-    }
-
-
-def _minute_frame_for_tplus1() -> pd.DataFrame:
-    minute = pd.read_parquet(NIFTY_1M_PATH)
-    minute = minute.copy()
-    minute["dt"] = pd.to_datetime(minute["date"], errors="coerce")
-    for col in ("open", "high", "low", "close", "volume"):
-        if col in minute.columns:
-            minute[col] = pd.to_numeric(minute[col], errors="coerce")
-    minute = minute.dropna(subset=["dt", "open", "high", "low", "close"]).sort_values("dt").reset_index(drop=True)
-    minute["session_date"] = minute["dt"].dt.normalize()
-    minute["time"] = minute["dt"].dt.strftime("%H:%M")
-    return minute
-
-
-def _build_tplus1_session_features(minute: pd.DataFrame) -> pd.DataFrame:
-    window = minute[(minute["time"] >= "14:00") & (minute["time"] <= "14:20")].copy()
-    window["minute_offset"] = window.groupby("session_date", sort=True).cumcount()
-    grouped = window.groupby("session_date", sort=True)
-    base = grouped.agg(
-        window_start=("dt", "first"),
-        window_end=("dt", "last"),
-        window_rows=("close", "size"),
-        w_open=("open", "first"),
-        w_high=("high", "max"),
-        w_low=("low", "min"),
-        w_close=("close", "last"),
-        w_volume=("volume", "sum") if "volume" in window.columns else ("close", "size"),
-    ).reset_index().rename(columns={"session_date": "date"})
-    base = base[base["window_rows"] == 21].copy()
-    base["w_return"] = safe_div(base["w_close"] - base["w_open"], base["w_open"])
-    base["w_range"] = safe_div(base["w_high"] - base["w_low"], base["w_open"])
-    base["w_body_to_range"] = safe_div(base["w_close"] - base["w_open"], base["w_high"] - base["w_low"])
-    base["w_close_location"] = safe_div(base["w_close"] - base["w_low"], base["w_high"] - base["w_low"])
-    window["ret_1m"] = window.groupby("session_date")["close"].pct_change(fill_method=None)
-    window["range_1m"] = safe_div(window["high"] - window["low"], window["open"])
-    window["body_1m"] = safe_div(window["close"] - window["open"], window["open"])
-    minute_features = window.pivot(
-        index="session_date",
-        columns="minute_offset",
-        values=["open", "high", "low", "close", "ret_1m", "range_1m", "body_1m"],
-    )
-    minute_features.columns = [f"m{int(offset):02d}_{field}" for field, offset in minute_features.columns]
-    minute_features = minute_features.reset_index().rename(columns={"session_date": "date"})
-    session_close = (
-        minute.groupby("session_date", sort=True)
-        .agg(day_close=("close", "last"))
-        .reset_index()
-        .rename(columns={"session_date": "date"})
-    )
-    frame = base.merge(minute_features, on="date", how="left").merge(session_close, on="date", how="left")
-    for offset in range(21):
-        close_col = f"m{offset:02d}_close"
-        open_col = f"m{offset:02d}_open"
-        if close_col in frame.columns:
-            frame[f"m{offset:02d}_close_vs_window_open"] = safe_div(frame[close_col] - frame["w_open"], frame["w_open"])
-        if open_col in frame.columns and close_col in frame.columns:
-            frame[f"m{offset:02d}_close_vs_minute_open"] = safe_div(frame[close_col] - frame[open_col], frame[open_col])
-    frame["ret_first_5m"] = safe_div(frame["m04_close"] - frame["m00_open"], frame["m00_open"])
-    frame["ret_last_5m"] = safe_div(frame["m20_close"] - frame["m16_open"], frame["m16_open"])
-    frame["ret_mid_11m"] = safe_div(frame["m15_close"] - frame["m05_open"], frame["m05_open"])
-    frame["last5_minus_first5"] = frame["ret_last_5m"] - frame["ret_first_5m"]
-    frame["abs_window_return"] = frame["w_return"].abs()
-    frame["dow"] = frame["date"].dt.dayofweek
-    frame["dom"] = frame["date"].dt.day
-    frame["month"] = frame["date"].dt.month
-    return frame.sort_values("date").reset_index(drop=True)
-
-
-def _add_tplus1_target_features(features: pd.DataFrame) -> pd.DataFrame:
-    frame = features.copy()
-    frame["target_date"] = frame["date"].shift(-1)
-    frame["target_close"] = frame["day_close"].shift(-1)
-    frame["target_return_from_1420"] = safe_div(frame["target_close"] - frame["w_close"], frame["w_close"])
-    frame["target"] = (frame["target_return_from_1420"] > 0).astype("float64")
-    frame.loc[frame["target_close"].isna(), "target"] = np.nan
-    for lag in (1, 2, 3, 5, 10):
-        frame[f"prev_target_lag{lag}"] = frame["target"].shift(lag)
-        frame[f"prev_target_return_lag{lag}"] = frame["target_return_from_1420"].shift(lag)
-    for window in (3, 5, 10, 20, 40):
-        min_periods = max(2, window // 2)
-        frame[f"prev_target_mean{window}"] = frame["target"].shift(1).rolling(window, min_periods=min_periods).mean()
-        shifted_return = frame["target_return_from_1420"].shift(1)
-        frame[f"prev_target_return_mean{window}"] = shifted_return.rolling(window, min_periods=min_periods).mean()
-        frame[f"prev_target_return_std{window}"] = shifted_return.rolling(window, min_periods=min_periods).std()
-    return frame
-
-
-def _apply_tplus1_overlays(pred: np.ndarray, frame: pd.DataFrame, overlays: list[dict[str, Any]]) -> np.ndarray:
-    adjusted = np.asarray(pred, dtype="int64").copy()
-    for overlay in overlays:
-        feature = str(overlay.get("feature", ""))
-        if feature not in frame.columns:
-            continue
-        series = pd.to_numeric(frame[feature], errors="coerce")
-        value = float(overlay.get("value", 0.0))
-        if overlay.get("op") == "<=":
-            mask = (series <= value).fillna(False).to_numpy(dtype=bool)
-        else:
-            mask = (series >= value).fillna(False).to_numpy(dtype=bool)
-        action = overlay.get("action")
-        if action == "up":
-            adjusted[mask] = 1
-        elif action == "down":
-            adjusted[mask] = 0
-        elif action == "flip":
-            adjusted[mask] = 1 - adjusted[mask]
-    return adjusted
-
-
-def refresh_tplus1_prediction(session_date: date | None = None) -> dict[str, Any]:
-    if not TPLUS1_MODEL_PATH.exists():
-        raise FileNotFoundError(f"Missing T+1 model artifact: {TPLUS1_MODEL_PATH}")
-    payload = joblib.load(TPLUS1_MODEL_PATH)
-    features = payload["features"]
-    threshold = float(payload["threshold"])
-    frame = _add_tplus1_target_features(_build_tplus1_session_features(_minute_frame_for_tplus1()))
-    if session_date is not None:
-        row = frame[pd.to_datetime(frame["date"], errors="coerce").dt.date == session_date].tail(1)
-    else:
-        row = frame.tail(1)
-    if row.empty:
-        minutes = fetch_yahoo_minutes(period="7d")
-        append_parquet_rows(NIFTY_1M_PATH, minutes, ["date"])
-        frame = _add_tplus1_target_features(_build_tplus1_session_features(_minute_frame_for_tplus1()))
-        if session_date is not None:
-            row = frame[pd.to_datetime(frame["date"], errors="coerce").dt.date == session_date].tail(1)
-        else:
-            row = frame.tail(1)
-    if row.empty:
-        raise RuntimeError("No complete 14:00-14:20 window is available for T+1 prediction.")
-    missing = [col for col in features if col not in row.columns]
-    if missing:
-        raise RuntimeError(f"T+1 feature row is missing model features: {missing[:5]}")
-    prob_up = predict_proba_up(payload["model"], row[features])
-    raw_pred = (prob_up >= threshold).astype("int64")
-    overlay_payload = payload.get("decision_overlay")
-    overlays = overlay_payload.get("overlays", []) if isinstance(overlay_payload, dict) else []
-    pred_int = int(_apply_tplus1_overlays(raw_pred, row, overlays)[0])
-    prediction = "UP" if pred_int == 1 else "DOWN"
-    input_day = pd.to_datetime(row["date"].iloc[0]).date()
-    target_day = next_trading_day(input_day + timedelta(days=1))
-    summary = load_tplus1_summary()
-    out = {
-        "input_date": input_day.isoformat(),
-        "target_date": target_day.isoformat(),
-        "forecast_for": f"next trading session after {input_day.isoformat()}",
-        "prediction": prediction,
-        "prob_up": float(prob_up[0]),
-        "confidence": float(max(prob_up[0], 1.0 - prob_up[0])),
-        "threshold": threshold,
-        "model_name": str(payload.get("model_name", summary.get("model_name", "nifty_1420_tplus1_logistic_model"))),
-        "decision_overlay": summary.get("decision_overlay"),
-        "validation_accuracy": summary.get("validation_accuracy"),
-        "test_accuracy": summary.get("test_accuracy"),
-        "accuracy_goal": summary.get("accuracy_goal"),
-    }
-    pd.DataFrame([out]).to_csv(TPLUS1_LATEST_PATH, index=False)
-    clear_dashboard_payload_cache()
-    return out
-
-
-def _tomorrow_probability_from_daily(daily: pd.DataFrame, fallback_prob: float) -> float:
-    if daily.empty or len(daily) < 5:
-        return float(fallback_prob)
-    frame = daily.copy()
-    frame["close"] = pd.to_numeric(frame["close"], errors="coerce")
-    frame = frame.dropna(subset=["close"]).tail(20)
-    if len(frame) < 5:
-        return float(fallback_prob)
-    close = frame["close"]
-    ret_1 = close.pct_change(fill_method=None).iloc[-1]
-    ret_5 = close.pct_change(5, fill_method=None).iloc[-1]
-    vol = close.pct_change(fill_method=None).tail(10).std()
-    score = 0.49900560447008563
-    if pd.notna(ret_1):
-        score += float(np.clip(ret_1 * 4.5, -0.05, 0.05))
-    if pd.notna(ret_5):
-        score += float(np.clip(ret_5 * 1.4, -0.05, 0.05))
-    if pd.notna(vol):
-        score -= float(np.clip(vol * 0.9, 0.0, 0.035))
-    return float(np.clip(score, 0.35, 0.65))
-
-
-def refresh_tomorrow_prediction(session_date: date | None = None) -> dict[str, Any]:
-    synced = sync_daily_forecaster_outputs()
-    if synced is not None and TOMORROW_LATEST_PATH.exists():
-        latest = pd.read_csv(TOMORROW_LATEST_PATH).iloc[-1].to_dict()
-        cleaned = {k: (None if pd.isna(v) else v) for k, v in latest.items()}
-        if session_date is None:
-            clear_dashboard_payload_cache()
-            return cleaned
-        try:
-            input_day = date.fromisoformat(str(cleaned.get("input_date"))[:10])
-        except Exception:
-            input_day = None
-        if input_day is not None:
-            clear_dashboard_payload_cache()
-            return cleaned
-    summary = load_tomorrow_summary()
-    artifact = load_tomorrow_model_artifact()
-    daily = pd.read_parquet(NIFTY_1D_PATH)
-    daily["date"] = pd.to_datetime(daily["date"], errors="coerce").dt.normalize()
-    daily = daily.dropna(subset=["date"]).sort_values("date")
-    if daily.empty:
-        raise RuntimeError("No daily NIFTY rows are available for tomorrow forecast.")
-    input_day = session_date or daily["date"].max().date()
-    target_day = next_trading_day(input_day + timedelta(days=1))
-    threshold = float(artifact.get("threshold", summary.get("threshold", 0.543)))
-    fallback_prob = float(summary.get("latest_forecast_prob_up", 0.49900560447008563))
-    prob_up = _tomorrow_probability_from_daily(daily[daily["date"].dt.date <= input_day], fallback_prob)
-    prediction = "UP" if prob_up >= threshold else "DOWN"
-    confidence = float(max(prob_up, 1.0 - prob_up))
-    row = {
-        "input_date": input_day.isoformat(),
-        "target_date": target_day.isoformat(),
-        "prediction": prediction,
-        "prob_up": prob_up,
-        "confidence": confidence,
-        "threshold": threshold,
-        "model_name": str(summary.get("model_name", "nifty_tomorrow_direction_model")),
-        "source_model": str(summary.get("source_model", "tuned_daily_forest_single")),
-        "validation_accuracy": float(summary.get("validation_accuracy", 0.5780141843971631)),
-        "test_accuracy": float(summary.get("test_accuracy", 0.6182795698924731)),
-    }
-    pd.DataFrame([row]).to_csv(TOMORROW_LATEST_PATH, index=False)
-    summary = dict(summary)
-    summary.update(
-        {
-            "latest_forecast_date": row["input_date"],
-            "latest_forecast_for": f"next trading session {row['target_date']}",
-            "latest_forecast_prob_up": row["prob_up"],
-            "latest_forecast_signal": row["prediction"],
-            "latest_target_date": row["target_date"],
-        }
-    )
-    TOMORROW_SUMMARY_PATH.write_text(json.dumps(summary, indent=2), encoding="utf-8")
-    clear_dashboard_payload_cache()
-    return row
-
-
-def _json_ready_frame(df: pd.DataFrame, limit: int | None = None) -> list[dict[str, Any]]:
-    out = df.copy()
-    if limit is not None:
-        out = out.tail(limit)
-    for col in out.columns:
-        if pd.api.types.is_datetime64_any_dtype(out[col]):
-            out[col] = out[col].dt.strftime("%Y-%m-%d %H:%M:%S")
-    out = out.replace({np.nan: None})
-    return out.to_dict(orient="records")
-
-
-def load_model_summary() -> dict[str, Any]:
-    summary_path = MODEL_DIR / "summary.json"
-    if not summary_path.exists():
-        return {}
-    return json.loads(summary_path.read_text(encoding="utf-8"))
-
-
-def load_candidate_results() -> list[dict[str, Any]]:
-    path = MODEL_DIR / "candidate_results.csv"
-    if not path.exists():
-        return []
-    return _json_ready_frame(pd.read_csv(path).head(12))
-
-
-def load_test_predictions() -> pd.DataFrame:
-    if not TEST_PREDICTIONS_PATH.exists():
-        return pd.DataFrame()
-    df = pd.read_parquet(TEST_PREDICTIONS_PATH)
-    df["date"] = pd.to_datetime(df["date"], errors="coerce")
-    return df.sort_values("date").reset_index(drop=True)
-
-
-def load_tomorrow_test_predictions() -> pd.DataFrame:
-    if not TOMORROW_TEST_PREDICTIONS_PATH.exists():
-        return pd.DataFrame()
-    df = pd.read_parquet(TOMORROW_TEST_PREDICTIONS_PATH)
-    for col in ("forecast_date", "target_date", "date"):
-        if col in df.columns:
-            df[col] = pd.to_datetime(df[col], errors="coerce")
-    sort_col = "target_date" if "target_date" in df.columns else "forecast_date"
-    return df.sort_values(sort_col).reset_index(drop=True)
-
-
-def load_tplus1_test_predictions() -> pd.DataFrame:
-    if not TPLUS1_TEST_PREDICTIONS_PATH.exists():
-        return pd.DataFrame()
-    df = pd.read_parquet(TPLUS1_TEST_PREDICTIONS_PATH)
-    for col in ("date", "target_date"):
-        if col in df.columns:
-            df[col] = pd.to_datetime(df[col], errors="coerce")
-    return df.sort_values("date").reset_index(drop=True)
-
-
-def dashboard_payload() -> dict[str, Any]:
-    key = (
-        _file_cache_key(MODEL_DIR / "summary.json"),
-        _file_cache_key(LATEST_PATH),
-        _file_cache_key(TEST_PREDICTIONS_PATH),
-        _file_cache_key(TOMORROW_SUMMARY_PATH),
-        _file_cache_key(TOMORROW_LATEST_PATH),
-        _file_cache_key(TOMORROW_TEST_PREDICTIONS_PATH),
-        _file_cache_key(TOMORROW_MODEL_PATH),
-        _file_cache_key(TPLUS1_SUMMARY_PATH),
-        _file_cache_key(TPLUS1_LATEST_PATH),
-        _file_cache_key(TPLUS1_TEST_PREDICTIONS_PATH),
-        _file_cache_key(TPLUS1_MODEL_PATH),
-        _file_cache_key(REFRESH_STATE_PATH),
-        _file_cache_key(NIFTY_1D_PATH),
-        _file_cache_key(OPENING_DATASET_PATH),
-        _file_cache_key(MODEL_DIR / "candidate_results.csv"),
-        _file_cache_key(NIFTY_1M_PATH),
-        _file_cache_key(LIVE_ACCURACY_PATH),
-    )
-    with _dashboard_payload_lock:
-        return copy.deepcopy(_dashboard_payload_cached(key))
-
-
-def warm_dashboard_payload_cache() -> None:
-    dashboard_payload()
-
-
-@lru_cache(maxsize=4)
-def _dashboard_payload_cached(key: tuple[tuple[str, int | None, int | None], ...]) -> dict[str, Any]:
-    summary = load_model_summary()
-    t5_latest = _latest_saved_prediction_uncached()
-    tomorrow_summary = load_tomorrow_summary()
-    tomorrow_latest = latest_tomorrow_prediction()
-    tplus1_summary = load_tplus1_summary()
-    tplus1_latest = latest_tplus1_prediction()
-    refresh_state = load_refresh_state()
-    t5_test = load_test_predictions()
-    tomorrow_test = load_tomorrow_test_predictions()
-    tplus1_test = load_tplus1_test_predictions()
-    daily = pd.read_parquet(NIFTY_1D_PATH)
-    daily["date"] = pd.to_datetime(daily["date"], errors="coerce")
-    daily = daily.sort_values("date").tail(180)
-    dataset = read_training_dataset()
-    opening = dataset[["date", "first5_return", "first5_range_pct", "first5_close_location"]].tail(120).copy()
-
-    if not t5_test.empty:
-        recent_predictions = t5_test.tail(40).copy()
-        recent_accuracy = float(recent_predictions["correct"].mean())
-        direction_mix = t5_test.groupby("prediction")["correct"].agg(["count", "mean"]).reset_index()
-        monthly = (
-            t5_test.assign(month=t5_test["date"].dt.strftime("%Y-%m"))
-            .groupby("month", as_index=False)["correct"]
-            .mean()
-            .rename(columns={"correct": "accuracy"})
-        )
-    else:
-        recent_predictions = pd.DataFrame()
-        recent_accuracy = None
-        direction_mix = pd.DataFrame()
-        monthly = pd.DataFrame()
-
-    if not tomorrow_test.empty:
-        tomorrow_recent = tomorrow_test.tail(40).copy()
-        if "pred" in tomorrow_recent.columns and "prediction" not in tomorrow_recent.columns:
-            tomorrow_recent["prediction"] = np.where(pd.to_numeric(tomorrow_recent["pred"], errors="coerce") == 1, "UP", "DOWN")
-        if "correct" not in tomorrow_recent.columns and {"target", "pred"}.issubset(tomorrow_recent.columns):
-            tomorrow_recent["correct"] = pd.to_numeric(tomorrow_recent["target"], errors="coerce") == pd.to_numeric(tomorrow_recent["pred"], errors="coerce")
-        tomorrow_accuracy = float(tomorrow_recent["correct"].mean()) if "correct" in tomorrow_recent.columns else tomorrow_summary.get("test_accuracy")
-    else:
-        tomorrow_recent = pd.DataFrame()
-        tomorrow_accuracy = tomorrow_summary.get("test_accuracy")
-
-    model_metrics = [
-        {
-            "id": "tomorrow",
-            "label": "Tomorrow",
-            "model_name": tomorrow_summary.get("model_name", "nifty_tomorrow_direction_model"),
-            "source_model": tomorrow_summary.get("source_model", "tuned_daily_forest_single"),
-            "validation_accuracy": tomorrow_summary.get("validation_accuracy"),
-            "test_accuracy": tomorrow_summary.get("test_accuracy"),
-            "recent_accuracy": tomorrow_accuracy,
-            "test_rows": int(tomorrow_summary.get("n_test") or len(tomorrow_test) or 0),
-        },
-        {
-            "id": "tplus1",
-            "label": "T+1",
-            "model_name": tplus1_summary.get("model_name", "nifty_1420_tplus1_logistic_model"),
-            "source_model": "14:00-14:20 logistic forecaster",
-            "validation_accuracy": tplus1_summary.get("validation_accuracy"),
-            "test_accuracy": tplus1_summary.get("test_accuracy"),
-            "recent_accuracy": float(tplus1_test.tail(40)["correct"].mean()) if not tplus1_test.empty and "correct" in tplus1_test.columns else tplus1_summary.get("test_accuracy"),
-            "test_rows": int(tplus1_summary.get("test_rows") or len(tplus1_test) or 0),
-        },
-        {
-            "id": "t5",
-            "label": "T+5",
-            "model_name": summary.get("model_name", "nifty_opening_direction_model"),
-            "source_model": summary.get("model_name", "nifty_opening_direction_model"),
-            "validation_accuracy": summary.get("validation_accuracy"),
-            "test_accuracy": summary.get("test_accuracy"),
-            "recent_accuracy": recent_accuracy,
-            "test_rows": int(len(t5_test)) if not t5_test.empty else int(summary.get("test_rows") or 0),
-        },
-    ]
-    metrics = {
-        "validation_accuracy": tomorrow_summary.get("validation_accuracy"),
-        "test_accuracy": tomorrow_summary.get("test_accuracy"),
-        "baseline_test_accuracy": tomorrow_summary.get("baseline_accuracy"),
-        "validation_auc": summary.get("validation_auc"),
-        "test_auc": summary.get("test_auc"),
-        "test_brier": summary.get("test_brier"),
-        "feature_count": tomorrow_summary.get("feature_count"),
-        "recent_accuracy": tomorrow_accuracy,
-        "recent_accuracy_days": int(len(tomorrow_recent)) if not tomorrow_recent.empty else 0,
-        "total_test_days": int(tomorrow_summary.get("n_test") or len(tomorrow_test) or 0),
-        "models": model_metrics,
-    }
-    return {
-        "latest": t5_latest,
-        "tomorrow_latest": tomorrow_latest,
-        "tplus1_latest": tplus1_latest,
-        "live_accuracy": load_live_accuracy(),
-        "metrics": metrics,
-        "summary": summary,
-        "tomorrow_summary": tomorrow_summary,
-        "tplus1_summary": tplus1_summary,
-        "candidates": load_candidate_results(),
-        "charts": {
-            "daily_close": _json_ready_frame(daily[["date", "open", "high", "low", "close"]]),
-            "opening_features": _json_ready_frame(opening),
-            "monthly_accuracy": _json_ready_frame(monthly),
-            "direction_mix": _json_ready_frame(direction_mix),
-            "recent_predictions": _json_ready_frame(recent_predictions),
-            "t5_recent_predictions": _json_ready_frame(recent_predictions),
-            "tomorrow_recent_predictions": _json_ready_frame(tomorrow_recent),
-            "tplus1_recent_predictions": _json_ready_frame(tplus1_test.tail(40)),
-        },
-        "data_status": {
-            "nifty_1m_rows": int(len(pd.read_parquet(NIFTY_1M_PATH, columns=["date"]))),
-            "nifty_1d_rows": int(len(pd.read_parquet(NIFTY_1D_PATH, columns=["date"]))),
-            "training_rows": int(len(dataset)),
-            "test_prediction_rows": int(len(t5_test)),
-            "tomorrow_test_prediction_rows": int(len(tomorrow_test)),
-            "tplus1_test_prediction_rows": int(len(tplus1_test)),
-            "latest_daily_date": pd.to_datetime(daily["date"]).max().date().isoformat(),
-            "refresh_phase": refresh_state.get("phase", REFRESH_NORMAL),
-            "refresh_state": refresh_state,
-        },
-    }
-
-
-def refresh_first5_prediction(session_date: date | None = None, minutes: pd.DataFrame | None = None) -> Prediction:
-    if session_date is None:
-        today = datetime.now(IST).date()
-        if not is_trading_day(today):
-            raise RuntimeError(f"{today.isoformat()} is not an NSE trading session.")
-    minutes = fetch_yahoo_minutes(period="7d") if minutes is None else minutes
-    append_parquet_rows(NIFTY_1M_PATH, minutes, ["date"])
-    first5 = first5_features_from_minutes(minutes, session_date=session_date)
-    row = build_model_row(first5)
-    dataset = read_training_dataset()
-    merged = pd.concat([dataset, row], ignore_index=True)
-    merged = merged.drop_duplicates(subset=["date"], keep="last").sort_values("date").reset_index(drop=True)
-    merged.to_parquet(OPENING_DATASET_PATH, index=False, compression="zstd")
-    prediction = predict_row(row)
-    clear_dashboard_payload_cache()
-    return prediction
-
-
-def refresh_daily_data() -> dict[str, Any]:
-    daily = fetch_yahoo_daily(period="1mo")
-    combined = append_parquet_rows(NIFTY_1D_PATH, daily, ["date"])
-    clear_dashboard_payload_cache()
-    return {
-        "rows": int(len(combined)),
-        "latest_date": pd.to_datetime(combined["date"]).max().date().isoformat(),
-        "path": str(NIFTY_1D_PATH),
-    }
-
-
-def update_opening_outcomes_from_daily() -> dict[str, Any]:
-    if not OPENING_DATASET_PATH.exists() or not NIFTY_1D_PATH.exists():
-        return {"updated_rows": 0, "latest_date": None}
-    dataset = pd.read_parquet(OPENING_DATASET_PATH)
-    daily = pd.read_parquet(NIFTY_1D_PATH)
-    if dataset.empty or daily.empty:
-        return {"updated_rows": 0, "latest_date": None}
-
-    dataset = dataset.copy()
-    dataset["_session_date"] = pd.to_datetime(dataset["date"], errors="coerce").dt.normalize()
-    daily = daily.copy()
-    daily["_session_date"] = pd.to_datetime(daily["date"], errors="coerce").dt.normalize()
-    daily = daily.dropna(subset=["_session_date"]).drop_duplicates("_session_date", keep="last")
-    daily = daily.set_index("_session_date")
-
-    updated = 0
-    for idx, session_day in dataset["_session_date"].dropna().items():
-        if session_day not in daily.index:
-            continue
-        row = daily.loc[session_day]
-        for src, dst in (
-            ("open", "day_open"),
-            ("high", "day_high"),
-            ("low", "day_low"),
-            ("close", "day_close"),
-            ("volume", "day_volume"),
-        ):
-            if src in row.index and dst in dataset.columns:
-                dataset.at[idx, dst] = row[src]
-        if {"day_open", "day_close", "target", "day_return"}.issubset(dataset.columns):
-            day_open = dataset.at[idx, "day_open"]
-            day_close = dataset.at[idx, "day_close"]
-            if pd.notna(day_open) and pd.notna(day_close) and float(day_open) != 0.0:
-                dataset.at[idx, "target"] = int(float(day_close) > float(day_open))
-                dataset.at[idx, "day_return"] = (float(day_close) - float(day_open)) / float(day_open)
-                updated += 1
-        if {"first5_close", "day_open", "first5_vs_day_open"}.issubset(dataset.columns):
-            first5_close = dataset.at[idx, "first5_close"]
-            day_open = dataset.at[idx, "day_open"]
-            if pd.notna(first5_close) and pd.notna(day_open) and float(day_open) != 0.0:
-                dataset.at[idx, "first5_vs_day_open"] = (float(first5_close) - float(day_open)) / float(day_open)
-
-    dataset = dataset.drop(columns=["_session_date"])
-    dataset = dataset.sort_values("date").reset_index(drop=True)
-    dataset.to_parquet(OPENING_DATASET_PATH, index=False, compression="zstd")
-    clear_dashboard_payload_cache()
-    latest = pd.to_datetime(dataset["date"], errors="coerce").max()
-    return {
-        "updated_rows": int(updated),
-        "latest_date": None if pd.isna(latest) else latest.date().isoformat(),
-    }
-
-
-def load_live_accuracy() -> dict[str, Any]:
-    """Load the live accuracy ledger from disk."""
-    if LIVE_ACCURACY_PATH.exists():
-        try:
-            return json.loads(LIVE_ACCURACY_PATH.read_text(encoding="utf-8"))
-        except Exception:
-            pass
-    return {
-        "tomorrow": {"entries": [], "accuracy": None, "total": 0, "correct_count": 0},
-        "t5": {"entries": [], "accuracy": None, "total": 0, "correct_count": 0},
-        "tplus1": {"entries": [], "accuracy": None, "total": 0, "correct_count": 0},
-    }
-
-
-def save_live_accuracy(data: dict[str, Any]) -> None:
-    """Persist the live accuracy ledger to disk."""
-    LIVE_ACCURACY_PATH.write_text(json.dumps(data, indent=2), encoding="utf-8")
-
-
-def update_live_accuracy(session_date: date) -> dict[str, Any]:
-    """Score today's predictions against actual outcomes and update the ledger.
-
-    Must be called AFTER refresh_daily_data() (so today's close is available)
-    but BEFORE refresh_first5_prediction / refresh_tplus1_prediction /
-    refresh_tomorrow_prediction (so the CSV files still hold the predictions
-    we want to score).
-    """
-    ledger = load_live_accuracy()
-    daily = pd.read_parquet(NIFTY_1D_PATH)
-    daily["_date"] = pd.to_datetime(daily["date"], errors="coerce").dt.normalize()
-    today_rows = daily[daily["_date"].dt.date == session_date]
-    if today_rows.empty:
-        return ledger
-
-    day_open = float(today_rows.iloc[-1]["open"])
-    day_close = float(today_rows.iloc[-1]["close"])
-    if not (np.isfinite(day_open) and np.isfinite(day_close) and day_open != 0):
-        return ledger
-    actual_close_gt_open = "UP" if day_close > day_open else "DOWN"
-    session_iso = session_date.isoformat()
-
-    # --- T+5: today's 9:20 AM prediction vs close > open ---
-    logged_t5 = {e["date"] for e in ledger["t5"]["entries"]}
-    if session_iso not in logged_t5 and LATEST_PATH.exists():
-        try:
-            t5_row = pd.read_csv(LATEST_PATH).iloc[-1].to_dict()
-            if str(t5_row.get("input_date", ""))[:10] == session_iso:
-                pred = str(t5_row.get("prediction", "")).upper()
-                if pred in ("UP", "DOWN"):
-                    ledger["t5"]["entries"].append({
-                        "date": session_iso,
-                        "prediction": pred,
-                        "actual": actual_close_gt_open,
-                        "correct": pred == actual_close_gt_open,
-                    })
-        except Exception:
-            pass
-
-    # --- Tomorrow: yesterday's prediction targeting today vs close > open ---
-    logged_tom = {e["date"] for e in ledger["tomorrow"]["entries"]}
-    if session_iso not in logged_tom and TOMORROW_LATEST_PATH.exists():
-        try:
-            tom_row = pd.read_csv(TOMORROW_LATEST_PATH).iloc[-1].to_dict()
-            if str(tom_row.get("target_date", ""))[:10] == session_iso:
-                pred = str(tom_row.get("prediction", "")).upper()
-                if pred in ("UP", "DOWN"):
-                    ledger["tomorrow"]["entries"].append({
-                        "date": session_iso,
-                        "prediction": pred,
-                        "actual": actual_close_gt_open,
-                        "correct": pred == actual_close_gt_open,
-                    })
-        except Exception:
-            pass
-
-    # --- T+1: yesterday's 14:20 prediction targeting today ---
-    # T+1 target: today's close > yesterday's 14:20 close
-    logged_t1 = {e["date"] for e in ledger["tplus1"]["entries"]}
-    if session_iso not in logged_t1 and TPLUS1_LATEST_PATH.exists():
-        try:
-            t1_row = pd.read_csv(TPLUS1_LATEST_PATH).iloc[-1].to_dict()
-            if str(t1_row.get("target_date", ""))[:10] == session_iso:
-                pred = str(t1_row.get("prediction", "")).upper()
-                input_date_str = str(t1_row.get("input_date", ""))[:10]
-                input_day = date.fromisoformat(input_date_str)
-                # Read the 14:20 close from minute data for the input session
-                minute = pd.read_parquet(NIFTY_1M_PATH, columns=["date", "close"])
-                minute["dt"] = pd.to_datetime(minute["date"], errors="coerce")
-                minute = minute.dropna(subset=["dt"])
-                minute["session_date"] = minute["dt"].dt.normalize()
-                minute["time_str"] = minute["dt"].dt.strftime("%H:%M")
-                window = minute[
-                    (minute["session_date"].dt.date == input_day)
-                    & (minute["time_str"] >= "14:00")
-                    & (minute["time_str"] <= "14:20")
-                ].sort_values("dt")
-                if not window.empty and pred in ("UP", "DOWN"):
-                    w_close = float(window.iloc[-1]["close"])
-                    t1_actual = "UP" if day_close > w_close else "DOWN"
-                    ledger["tplus1"]["entries"].append({
-                        "date": session_iso,
-                        "prediction": pred,
-                        "actual": t1_actual,
-                        "correct": pred == t1_actual,
-                    })
-        except Exception:
-            pass
-
-    # Recompute summary stats
-    for model_id in ("t5", "tomorrow", "tplus1"):
-        entries = ledger[model_id]["entries"]
-        total = len(entries)
-        correct = sum(1 for e in entries if e.get("correct"))
-        ledger[model_id]["total"] = total
-        ledger[model_id]["correct_count"] = correct
-        ledger[model_id]["accuracy"] = correct / total if total > 0 else None
-
-    save_live_accuracy(ledger)
-    clear_dashboard_payload_cache()
-    return ledger
-
-
-def refresh_market_close_data(session_date: date | None = None) -> dict[str, Any]:
-    now = datetime.now(IST)
-    session_date = session_date or now.date()
-    if not is_trading_day(session_date):
-        raise RuntimeError(f"{session_date.isoformat()} is not an NSE trading session.")
-    save_refresh_state(REFRESH_WAITING, session_date=session_date)
-    try:
-        save_refresh_state(REFRESH_REFRESHING, session_date=session_date)
-        minutes = fetch_yahoo_minutes(period="7d")
-        minute_frame = append_parquet_rows(NIFTY_1M_PATH, minutes, ["date"])
-        daily_info = refresh_daily_data()
-        # Score live predictions BEFORE they get overwritten by fresh ones
-        try:
-            update_live_accuracy(session_date)
-        except Exception as exc:
-            print(f"[close-refresh] live accuracy update failed: {exc}", flush=True)
-        t5_prediction = refresh_first5_prediction(session_date=session_date, minutes=minutes)
-        tplus1_prediction = refresh_tplus1_prediction(session_date=session_date)
-        outcomes = update_opening_outcomes_from_daily()
-        tomorrow_prediction = refresh_tomorrow_prediction(session_date=session_date)
-        state = save_refresh_state(REFRESH_READY, session_date=session_date)
-        clear_dashboard_payload_cache()
-        return {
-            "session_date": session_date.isoformat(),
-            "nifty_1m_rows": int(len(minute_frame)),
-            "latest_minute": pd.to_datetime(minute_frame["date"], errors="coerce").max().isoformat(),
-            "daily": daily_info,
-            "opening_dataset": outcomes,
-            "t5_prediction": t5_prediction.to_dict(),
-            "tplus1_prediction": tplus1_prediction,
-            "tomorrow_prediction": tomorrow_prediction,
-            "refresh_state": state,
-        }
-    except Exception as exc:
-        save_refresh_state(REFRESH_FAILED, session_date=session_date, error=str(exc))
-        clear_dashboard_payload_cache()
-        raise
-
-
-def close_refresh_due(now: datetime | None = None) -> bool:
-    now = now or datetime.now(IST)
-    if not is_trading_day(now.date()) or now.time() < CLOSE_REFRESH_READY:
-        return False
-    latest_daily = latest_parquet_date(NIFTY_1D_PATH)
-    latest_minutes = latest_parquet_date(NIFTY_1M_PATH)
-    latest_opening = latest_parquet_date(OPENING_DATASET_PATH)
-    latest_opening_outcome = latest_opening_outcome_date()
-    tomorrow_latest = latest_tomorrow_prediction()
-    tomorrow_input = None
-    try:
-        if tomorrow_latest.get("input_date"):
-            tomorrow_input = date.fromisoformat(str(tomorrow_latest.get("input_date"))[:10])
-    except Exception:
-        tomorrow_input = None
-    return any(
-        latest != now.date()
-        for latest in (latest_daily, latest_minutes, latest_opening, latest_opening_outcome, tomorrow_input)
-    )
-
-
-def latest_prediction_input_date(path: Path) -> date | None:
-    if not path.exists():
-        return None
-    try:
-        frame = pd.read_csv(path, usecols=["input_date"])
-    except Exception:
-        return None
-    if frame.empty:
-        return None
-    value = pd.to_datetime(frame["input_date"], errors="coerce").max()
-    return None if pd.isna(value) else value.date()
-
-
-def latest_tomorrow_input_date() -> date | None:
-    try:
-        latest = latest_tomorrow_prediction()
-        raw = latest.get("input_date")
-        return date.fromisoformat(str(raw)[:10]) if raw else None
-    except Exception:
-        return None
-
-
-def expected_completed_daily_date(now: datetime | None = None) -> date:
-    now = now or datetime.now(IST)
-    if is_trading_day(now.date()) and now.time() < CLOSE_REFRESH_READY:
-        return previous_trading_day(now.date() - timedelta(days=1))
-    return previous_trading_day(now.date())
-
-
-def expected_minute_date(now: datetime | None = None) -> date:
-    now = now or datetime.now(IST)
-    if is_trading_day(now.date()) and now.time() >= FIRST5_READY:
-        return now.date()
-    return previous_trading_day(now.date() - timedelta(days=1))
-
-
-def expected_tplus1_date(now: datetime | None = None) -> date:
-    now = now or datetime.now(IST)
-    if is_trading_day(now.date()) and now.time() >= TPLUS1_READY:
-        return now.date()
-    return previous_trading_day(now.date() - timedelta(days=1))
-
-
-def is_stale(latest: date | None, expected: date) -> bool:
-    return latest is None or latest < expected
-
-
-def stale_data_status(now: datetime | None = None) -> dict[str, Any]:
-    now = now or datetime.now(IST)
-    expected_daily = expected_completed_daily_date(now)
-    expected_minutes = expected_minute_date(now)
-    expected_tplus1 = expected_tplus1_date(now)
-    latest_daily = latest_parquet_date(NIFTY_1D_PATH)
-    latest_minutes = latest_parquet_date(NIFTY_1M_PATH)
-    latest_t5 = latest_prediction_input_date(LATEST_PATH)
-    latest_tomorrow = latest_tomorrow_input_date()
-    latest_tplus1 = latest_prediction_input_date(TPLUS1_LATEST_PATH)
-    return {
-        "server_time_ist": now.isoformat(),
-        "expected_daily_date": expected_daily.isoformat(),
-        "expected_minute_date": expected_minutes.isoformat(),
-        "expected_tplus1_date": expected_tplus1.isoformat(),
-        "latest_daily_date": latest_daily.isoformat() if latest_daily else None,
-        "latest_minute_date": latest_minutes.isoformat() if latest_minutes else None,
-        "latest_t5_date": latest_t5.isoformat() if latest_t5 else None,
-        "latest_tomorrow_date": latest_tomorrow.isoformat() if latest_tomorrow else None,
-        "latest_tplus1_date": latest_tplus1.isoformat() if latest_tplus1 else None,
-        "daily_stale": is_stale(latest_daily, expected_daily),
-        "minutes_stale": is_stale(latest_minutes, expected_minutes),
-        "t5_stale": is_stale(latest_t5, expected_minutes),
-        "tomorrow_stale": is_stale(latest_tomorrow, expected_daily),
-        "tplus1_stale": is_stale(latest_tplus1, expected_tplus1),
-    }
-
-
-def refresh_stale_data_once(now: datetime | None = None) -> dict[str, Any]:
-    now = now or datetime.now(IST)
-    status = stale_data_status(now)
-    if not any(status[key] for key in ("daily_stale", "minutes_stale", "t5_stale", "tomorrow_stale", "tplus1_stale")):
-        return {"status": "fresh", **status, "actions": []}
-    if not _stale_refresh_lock.acquire(blocking=False):
-        return {"status": "skipped", "reason": "stale refresh already running", **status, "actions": []}
-
-    actions: list[dict[str, Any]] = []
-    try:
-        if status["minutes_stale"]:
-            minutes = fetch_yahoo_minutes(period="7d")
-            combined = append_parquet_rows(NIFTY_1M_PATH, minutes, ["date"])
-            actions.append(
-                {
-                    "name": "minutes",
-                    "rows": int(len(combined)),
-                    "latest_date": pd.to_datetime(combined["date"], errors="coerce").max().date().isoformat(),
-                }
-            )
-
-        if status["daily_stale"]:
-            daily_info = refresh_daily_data()
-            outcomes = update_opening_outcomes_from_daily()
-            actions.append({"name": "daily", **daily_info})
-            actions.append({"name": "opening_outcomes", **outcomes})
-
-        if status["daily_stale"] or status["tomorrow_stale"]:
-            try:
-                tomorrow = refresh_tomorrow_prediction(session_date=date.fromisoformat(status["expected_daily_date"]))
-                actions.append({"name": "tomorrow_prediction", "input_date": tomorrow.get("input_date")})
-            except Exception as exc:
-                actions.append({"name": "tomorrow_prediction", "error": str(exc)})
-
-        if status["t5_stale"] and is_trading_day(now.date()) and now.time() >= FIRST5_READY:
-            prediction = refresh_first5_prediction(session_date=now.date())
-            actions.append({"name": "t5_prediction", "input_date": prediction.input_date})
-
-        if status["tplus1_stale"] and is_trading_day(now.date()) and now.time() >= TPLUS1_READY:
-            prediction = refresh_tplus1_prediction(session_date=now.date())
-            actions.append({"name": "tplus1_prediction", "input_date": prediction.get("input_date")})
-
-        clear_dashboard_payload_cache()
-        refreshed_status = stale_data_status(datetime.now(IST))
-        return {"status": "refreshed", **refreshed_status, "actions": actions}
-    finally:
-        _stale_refresh_lock.release()
-
-
-def next_ist_run_at(run_time: time = time(9, 20), now: datetime | None = None) -> datetime:
-    now = now or datetime.now(IST)
-    target_day = now.date()
-    if now >= datetime.combine(target_day, run_time, tzinfo=IST):
-        target_day += timedelta(days=1)
-    target_day = next_trading_day(target_day)
-    return datetime.combine(target_day, run_time, tzinfo=IST)
-
-
-def seconds_until_next_ist_run(run_time: time = time(9, 20)) -> float:
-    now = datetime.now(IST)
-    target = next_ist_run_at(run_time, now=now)
-    return max(1.0, (target - now).total_seconds())

backend/nifty_backend/yahoo_history_client.py

@@ -1,445 +0,0 @@
-from __future__ import annotations
-
-import gzip
-import hashlib
-import json
-import sqlite3
-import threading
-import time
-from dataclasses import dataclass
-from datetime import datetime, timedelta
-from itertools import cycle
-from pathlib import Path
-from typing import Any
-from zoneinfo import ZoneInfo
-
-import pandas as pd
-import requests
-from requests.adapters import HTTPAdapter
-from urllib3.util.retry import Retry
-
-
-YAHOO_CHART_HOSTS = (
-    "https://query1.finance.yahoo.com",
-    "https://query2.finance.yahoo.com",
-)
-DEFAULT_HEADERS = {
-    "User-Agent": (
-        "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
-        "AppleWebKit/537.36 (KHTML, like Gecko) "
-        "Chrome/136.0 Safari/537.36"
-    ),
-    "Accept": "application/json,text/plain,*/*",
-    "Accept-Language": "en-US,en;q=0.9",
-    "Connection": "keep-alive",
-    "Origin": "https://finance.yahoo.com",
-    "Referer": "https://finance.yahoo.com/",
-}
-BAR_COLUMNS = ["timestamp", "open", "high", "low", "close", "adj_close", "volume", "dividend", "split_ratio"]
-
-
-@dataclass(frozen=True)
-class IntervalPolicy:
-    interval: str
-    chunk_days: int
-    min_chunk_days: int
-    retention_days: int | None
-
-
-INTERVAL_POLICIES: dict[str, IntervalPolicy] = {
-    "1m": IntervalPolicy("1m", chunk_days=7, min_chunk_days=1, retention_days=30),
-    "2m": IntervalPolicy("2m", chunk_days=60, min_chunk_days=2, retention_days=60),
-    "5m": IntervalPolicy("5m", chunk_days=60, min_chunk_days=5, retention_days=60),
-    "15m": IntervalPolicy("15m", chunk_days=60, min_chunk_days=5, retention_days=60),
-    "30m": IntervalPolicy("30m", chunk_days=60, min_chunk_days=5, retention_days=60),
-    "60m": IntervalPolicy("60m", chunk_days=60, min_chunk_days=5, retention_days=60),
-    "90m": IntervalPolicy("90m", chunk_days=60, min_chunk_days=5, retention_days=60),
-    "1h": IntervalPolicy("1h", chunk_days=60, min_chunk_days=5, retention_days=60),
-    "1d": IntervalPolicy("1d", chunk_days=3650, min_chunk_days=30, retention_days=None),
-    "5d": IntervalPolicy("5d", chunk_days=3650, min_chunk_days=30, retention_days=None),
-    "1wk": IntervalPolicy("1wk", chunk_days=3650, min_chunk_days=30, retention_days=None),
-    "1mo": IntervalPolicy("1mo", chunk_days=3650, min_chunk_days=30, retention_days=None),
-    "3mo": IntervalPolicy("3mo", chunk_days=3650, min_chunk_days=30, retention_days=None),
-}
-
-
-class YahooHistoryError(RuntimeError):
-    pass
-
-
-class YahooSymbolError(YahooHistoryError):
-    pass
-
-
-class YahooIntervalLimitError(YahooHistoryError):
-    pass
-
-
-class YahooRateLimitError(YahooHistoryError):
-    pass
-
-
-class SqliteResponseCache:
-    def __init__(self, path: Path) -> None:
-        self.path = path
-        self.path.parent.mkdir(parents=True, exist_ok=True)
-        self._lock = threading.Lock()
-        with self._connect() as connection:
-            connection.execute(
-                """
-                CREATE TABLE IF NOT EXISTS response_cache (
-                    cache_key TEXT PRIMARY KEY,
-                    fetched_at INTEGER NOT NULL,
-                    payload_gzip BLOB NOT NULL
-                )
-                """
-            )
-
-    def _connect(self) -> sqlite3.Connection:
-        connection = sqlite3.connect(self.path)
-        connection.execute("PRAGMA journal_mode=WAL")
-        connection.execute("PRAGMA synchronous=NORMAL")
-        return connection
-
-    def get(self, cache_key: str, ttl_seconds: int) -> dict[str, Any] | None:
-        with self._lock, self._connect() as connection:
-            row = connection.execute(
-                "SELECT fetched_at, payload_gzip FROM response_cache WHERE cache_key = ?",
-                (cache_key,),
-            ).fetchone()
-        if row is None:
-            return None
-        fetched_at, payload_gzip = row
-        if int(time.time()) - int(fetched_at) > ttl_seconds:
-            return None
-        return json.loads(gzip.decompress(payload_gzip).decode("utf-8"))
-
-    def set(self, cache_key: str, payload: dict[str, Any]) -> None:
-        packed = gzip.compress(json.dumps(payload, separators=(",", ":"), ensure_ascii=True).encode("utf-8"))
-        with self._lock, self._connect() as connection:
-            connection.execute(
-                """
-                INSERT INTO response_cache (cache_key, fetched_at, payload_gzip)
-                VALUES (?, ?, ?)
-                ON CONFLICT(cache_key) DO UPDATE SET
-                    fetched_at = excluded.fetched_at,
-                    payload_gzip = excluded.payload_gzip
-                """,
-                (cache_key, int(time.time()), packed),
-            )
-
-
-class RateLimiter:
-    def __init__(self, min_gap_seconds: float) -> None:
-        self.min_gap_seconds = max(0.0, float(min_gap_seconds))
-        self._lock = threading.Lock()
-        self._next_allowed = 0.0
-
-    def wait(self) -> None:
-        with self._lock:
-            delay = self._next_allowed - time.monotonic()
-            if delay > 0:
-                time.sleep(delay)
-            self._next_allowed = time.monotonic() + self.min_gap_seconds
-
-
-class YahooHistoryClient:
-    def __init__(
-        self,
-        *,
-        cache_path: Path,
-        timeout_seconds: float = 25.0,
-        min_request_gap_seconds: float = 0.35,
-        max_retries: int = 5,
-    ) -> None:
-        self.cache = SqliteResponseCache(cache_path)
-        self.timeout_seconds = timeout_seconds
-        self.rate_limiter = RateLimiter(min_request_gap_seconds)
-        self.host_cycle = cycle(YAHOO_CHART_HOSTS)
-        self.session = self._build_session(max_retries=max_retries)
-
-    def _build_session(self, *, max_retries: int) -> requests.Session:
-        retry = Retry(
-            total=max_retries,
-            connect=max_retries,
-            read=max_retries,
-            backoff_factor=0.8,
-            status_forcelist=(429, 500, 502, 503, 504),
-            allowed_methods=("GET",),
-            respect_retry_after_header=True,
-            raise_on_status=False,
-        )
-        adapter = HTTPAdapter(max_retries=retry, pool_connections=16, pool_maxsize=16)
-        session = requests.Session()
-        session.headers.update(DEFAULT_HEADERS)
-        session.mount("https://", adapter)
-        session.mount("http://", adapter)
-        return session
-
-    def fetch_history(
-        self,
-        symbol: str,
-        *,
-        interval: str,
-        start: str | datetime,
-        end: str | datetime,
-        include_prepost: bool = False,
-        adjust_ohlc: bool = False,
-    ) -> pd.DataFrame:
-        policy = self._interval_policy(interval)
-        start_dt = self._coerce_datetime(start, end_of_day=False)
-        end_dt = self._coerce_datetime(end, end_of_day=True)
-        if end_dt <= start_dt:
-            raise ValueError("end must be later than start")
-        self._validate_retention_window(policy=policy, start_dt=start_dt, end_dt=end_dt)
-
-        frames: list[pd.DataFrame] = []
-        for chunk_start, chunk_end in self._iter_chunks(start_dt=start_dt, end_dt=end_dt, chunk_days=policy.chunk_days):
-            chunk = self._fetch_chunk_adaptive(
-                symbol=symbol,
-                interval=policy.interval,
-                chunk_start=chunk_start,
-                chunk_end=chunk_end,
-                min_chunk_days=policy.min_chunk_days,
-                include_prepost=include_prepost,
-                adjust_ohlc=adjust_ohlc,
-            )
-            if not chunk.empty:
-                frames.append(chunk)
-        if not frames:
-            return pd.DataFrame(columns=BAR_COLUMNS)
-        history = pd.concat(frames, ignore_index=True)
-        history = history.drop_duplicates(subset=["timestamp"], keep="last").sort_values("timestamp").reset_index(drop=True)
-        return history[(history["timestamp"] >= start_dt) & (history["timestamp"] <= end_dt)].reset_index(drop=True)
-
-    def _fetch_chunk_adaptive(
-        self,
-        *,
-        symbol: str,
-        interval: str,
-        chunk_start: datetime,
-        chunk_end: datetime,
-        min_chunk_days: int,
-        include_prepost: bool,
-        adjust_ohlc: bool,
-    ) -> pd.DataFrame:
-        try:
-            payload = self._request_chart(
-                symbol=symbol,
-                interval=interval,
-                start_dt=chunk_start,
-                end_dt=chunk_end,
-                include_prepost=include_prepost,
-            )
-            return self._payload_to_frame(payload=payload, adjust_ohlc=adjust_ohlc)
-        except YahooIntervalLimitError:
-            if max((chunk_end - chunk_start).days, 1) <= min_chunk_days:
-                raise
-            midpoint = chunk_start + (chunk_end - chunk_start) / 2
-            left = self._fetch_chunk_adaptive(
-                symbol=symbol,
-                interval=interval,
-                chunk_start=chunk_start,
-                chunk_end=midpoint,
-                min_chunk_days=min_chunk_days,
-                include_prepost=include_prepost,
-                adjust_ohlc=adjust_ohlc,
-            )
-            right = self._fetch_chunk_adaptive(
-                symbol=symbol,
-                interval=interval,
-                chunk_start=midpoint,
-                chunk_end=chunk_end,
-                min_chunk_days=min_chunk_days,
-                include_prepost=include_prepost,
-                adjust_ohlc=adjust_ohlc,
-            )
-            return pd.concat([left, right], ignore_index=True)
-
-    def _request_chart(
-        self,
-        *,
-        symbol: str,
-        interval: str,
-        start_dt: datetime,
-        end_dt: datetime,
-        include_prepost: bool,
-    ) -> dict[str, Any]:
-        params = {
-            "period1": str(int(start_dt.timestamp())),
-            "period2": str(int(end_dt.timestamp())),
-            "interval": interval,
-            "includePrePost": "true" if include_prepost else "false",
-            "events": "div,splits,capitalGains",
-            "includeAdjustedClose": "true",
-        }
-        last_error: Exception | None = None
-        for _ in range(len(YAHOO_CHART_HOSTS)):
-            base_url = next(self.host_cycle)
-            url = f"{base_url}/v8/finance/chart/{requests.utils.quote(symbol, safe='')}"
-            cache_key = self._cache_key(url=url, params=params)
-            cached = self.cache.get(cache_key, ttl_seconds=self._cache_ttl_seconds(end_dt=end_dt))
-            if cached is not None:
-                return cached
-
-            self.rate_limiter.wait()
-            response = self.session.get(url, params=params, timeout=self.timeout_seconds)
-            if response.status_code == 429:
-                last_error = YahooRateLimitError(f"Yahoo rate-limited {symbol} at interval {interval}.")
-                time.sleep(1.5)
-                continue
-            if response.status_code == 404:
-                raise YahooSymbolError(f"Yahoo did not recognize ticker {symbol}.")
-            if response.status_code == 422:
-                raise YahooIntervalLimitError(
-                    f"Yahoo rejected {symbol} {interval} from {start_dt.isoformat()} to {end_dt.isoformat()}."
-                )
-            try:
-                response.raise_for_status()
-            except requests.HTTPError as exc:
-                last_error = exc
-                continue
-
-            payload = response.json()
-            error = payload.get("chart", {}).get("error")
-            if error:
-                description = error.get("description") or error.get("code") or str(error)
-                lowered = description.lower()
-                if "not found" in lowered or "no data found" in lowered or "symbol" in lowered:
-                    raise YahooSymbolError(description)
-                if "range" in lowered or "interval" in lowered or "last" in lowered:
-                    raise YahooIntervalLimitError(description)
-                if "rate limit" in lowered or "too many requests" in lowered:
-                    raise YahooRateLimitError(description)
-                raise YahooHistoryError(description)
-            self.cache.set(cache_key, payload)
-            return payload
-
-        if last_error is not None:
-            raise YahooHistoryError(str(last_error)) from last_error
-        raise YahooHistoryError(f"Yahoo request failed for {symbol} {interval}.")
-
-    def _payload_to_frame(self, *, payload: dict[str, Any], adjust_ohlc: bool) -> pd.DataFrame:
-        result = payload.get("chart", {}).get("result") or []
-        if not result:
-            return pd.DataFrame(columns=BAR_COLUMNS)
-        result0 = result[0]
-        meta = result0.get("meta") or {}
-        timestamps = result0.get("timestamp") or []
-        quote_sets = result0.get("indicators", {}).get("quote") or []
-        if not timestamps or not quote_sets:
-            return pd.DataFrame(columns=BAR_COLUMNS)
-
-        try:
-            timezone = ZoneInfo(meta.get("exchangeTimezoneName") or "UTC")
-        except Exception:
-            timezone = ZoneInfo("UTC")
-        quote = quote_sets[0]
-        adjclose_sets = result0.get("indicators", {}).get("adjclose") or [{}]
-        adj_close = adjclose_sets[0].get("adjclose", []) if adjclose_sets else []
-        events = result0.get("events") or {}
-        dividends = self._event_series(events.get("dividends") or {}, value_key="amount")
-        splits = self._event_series(events.get("splits") or {}, value_key="splitRatio")
-        row_count = len(timestamps)
-
-        frame = pd.DataFrame(
-            {
-                "timestamp": pd.to_datetime(timestamps, unit="s", utc=True).tz_convert(timezone).tz_localize(None),
-                "open": self._normalize_values(quote.get("open", []), row_count),
-                "high": self._normalize_values(quote.get("high", []), row_count),
-                "low": self._normalize_values(quote.get("low", []), row_count),
-                "close": self._normalize_values(quote.get("close", []), row_count),
-                "adj_close": self._normalize_values(adj_close, row_count),
-                "volume": self._normalize_values(quote.get("volume", []), row_count),
-            }
-        )
-        for column in ("open", "high", "low", "close", "adj_close", "volume"):
-            frame[column] = pd.to_numeric(frame[column], errors="coerce")
-        frame = frame.dropna(subset=["timestamp", "close"]).reset_index(drop=True)
-        frame["volume"] = frame["volume"].fillna(0.0)
-
-        frame["epoch"] = (frame["timestamp"].astype("int64") // 1_000_000_000).astype("int64")
-        frame["dividend"] = frame["epoch"].map(dividends).fillna(0.0)
-        frame["split_ratio"] = frame["epoch"].map(splits).fillna(1.0)
-        frame = frame.drop(columns=["epoch"])
-
-        if adjust_ohlc:
-            ratio = frame["adj_close"].where(frame["close"] != 0, frame["close"]) / frame["close"].replace(0, pd.NA)
-            ratio = ratio.fillna(1.0)
-            for column in ("open", "high", "low", "close"):
-                frame[column] = frame[column] * ratio
-        return frame.drop_duplicates(subset=["timestamp"], keep="last").sort_values("timestamp").reset_index(drop=True)[BAR_COLUMNS]
-
-    @staticmethod
-    def _event_series(events: dict[str, Any], *, value_key: str) -> dict[int, float]:
-        output: dict[int, float] = {}
-        for event in events.values():
-            timestamp = event.get("date")
-            value = event.get(value_key)
-            if timestamp is None or value is None:
-                continue
-            try:
-                output[int(timestamp)] = float(value)
-            except (TypeError, ValueError):
-                continue
-        return output
-
-    @staticmethod
-    def _normalize_values(values: list[Any] | tuple[Any, ...], size: int) -> list[Any]:
-        normalized = list(values[:size])
-        if len(normalized) < size:
-            normalized.extend([None] * (size - len(normalized)))
-        return normalized
-
-    @staticmethod
-    def _cache_key(*, url: str, params: dict[str, str]) -> str:
-        material = json.dumps({"url": url, "params": params}, sort_keys=True, separators=(",", ":"))
-        return hashlib.sha256(material.encode("utf-8")).hexdigest()
-
-    @staticmethod
-    def _cache_ttl_seconds(*, end_dt: datetime) -> int:
-        now_utc = datetime.utcnow()
-        if end_dt < now_utc - timedelta(days=2):
-            return 7 * 24 * 60 * 60
-        if end_dt < now_utc - timedelta(hours=12):
-            return 60 * 60
-        return 90
-
-    @staticmethod
-    def _coerce_datetime(value: str | datetime, *, end_of_day: bool) -> datetime:
-        timestamp = pd.Timestamp(value)
-        if timestamp.tzinfo is not None:
-            timestamp = timestamp.tz_convert("UTC").tz_localize(None)
-        dt = timestamp.to_pydatetime()
-        if end_of_day and dt.hour == 0 and dt.minute == 0 and dt.second == 0 and dt.microsecond == 0:
-            return dt + timedelta(days=1)
-        return dt
-
-    @staticmethod
-    def _iter_chunks(*, start_dt: datetime, end_dt: datetime, chunk_days: int) -> list[tuple[datetime, datetime]]:
-        chunks: list[tuple[datetime, datetime]] = []
-        cursor = start_dt
-        while cursor < end_dt:
-            next_edge = min(cursor + timedelta(days=chunk_days), end_dt)
-            chunks.append((cursor, next_edge))
-            cursor = next_edge
-        return chunks
-
-    @staticmethod
-    def _interval_policy(interval: str) -> IntervalPolicy:
-        normalized = interval.strip()
-        if normalized not in INTERVAL_POLICIES:
-            allowed = ", ".join(sorted(INTERVAL_POLICIES))
-            raise ValueError(f"Unsupported interval {interval!r}. Allowed values: {allowed}")
-        return INTERVAL_POLICIES[normalized]
-
-    @staticmethod
-    def _validate_retention_window(*, policy: IntervalPolicy, start_dt: datetime, end_dt: datetime) -> None:
-        if policy.retention_days is None:
-            return
-        earliest = datetime.utcnow() - timedelta(days=policy.retention_days)
-        if start_dt < earliest or end_dt < earliest:
-            cutoff = earliest.strftime("%Y-%m-%d")
-            raise YahooIntervalLimitError(
-                f"Yahoo only serves {policy.interval} history back to about {cutoff}. Use 1d or coarser for deeper history."
-            )

backend/requirements.txt

@@ -1,10 +0,0 @@
-pandas
-pandas_market_calendars
-pyarrow
-requests
-fastapi
-uvicorn
-joblib
-numpy
-scikit-learn
-catboost

backend/scripts/refresh_daily_data.py

@@ -1,15 +0,0 @@
-from __future__ import annotations
-
-import sys
-from pathlib import Path
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
-from nifty_backend.runtime import refresh_daily_data
-
-
-def main() -> None:
-    print(refresh_daily_data())
-
-
-if __name__ == "__main__":
-    main()

backend/scripts/refresh_first5_prediction.py

@@ -1,26 +0,0 @@
-from __future__ import annotations
-
-import argparse
-import sys
-from datetime import date
-from pathlib import Path
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
-from nifty_backend.runtime import refresh_first5_prediction
-
-
-def parse_args() -> argparse.Namespace:
-    parser = argparse.ArgumentParser(description="Fetch Yahoo Finance first five NIFTY minutes and refresh prediction.")
-    parser.add_argument("--date", default=None, help="Optional IST session date, YYYY-MM-DD.")
-    return parser.parse_args()
-
-
-def main() -> None:
-    args = parse_args()
-    session_date = date.fromisoformat(args.date) if args.date else None
-    prediction = refresh_first5_prediction(session_date=session_date)
-    print(prediction.to_dict())
-
-
-if __name__ == "__main__":
-    main()

backend/scripts/retrain_opening_model.py

@@ -1,182 +0,0 @@
-from __future__ import annotations
-
-import argparse
-import json
-import sys
-from dataclasses import asdict, dataclass
-from pathlib import Path
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
-
-import joblib
-import numpy as np
-import pandas as pd
-from sklearn.ensemble import ExtraTreesClassifier
-from sklearn.impute import SimpleImputer
-from sklearn.linear_model import LogisticRegression
-from sklearn.metrics import accuracy_score, brier_score_loss, log_loss, roc_auc_score
-from sklearn.pipeline import make_pipeline
-from sklearn.preprocessing import StandardScaler
-
-from nifty_backend.runtime import (
-    DECISION_OVERLAYS,
-    MODEL_DIR,
-    MODEL_PATH,
-    OPENING_DATASET_PATH,
-    ProbabilityBlend,
-    apply_decision_overlays,
-    directional_confidence,
-    predict_proba_up,
-)
-
-
-DEFAULT_TRAIN_END = pd.Timestamp("2023-12-31")
-DEFAULT_VALID_END = pd.Timestamp("2025-08-17")
-RANDOM_SEED = 42
-
-
-@dataclass
-class RetrainSummary:
-    model_name: str
-    threshold: float
-    train_rows: int
-    valid_rows: int
-    test_rows: int
-    validation_accuracy: float
-    test_accuracy: float
-    validation_auc: float
-    test_auc: float
-    test_brier: float
-    latest_prediction: str
-    latest_prob_up: float
-    latest_confidence: float
-    feature_count: int
-
-
-def feature_columns(frame: pd.DataFrame) -> list[str]:
-    excluded = {
-        "date",
-        "first5_start",
-        "first5_end",
-        "target",
-        "day_open",
-        "day_high",
-        "day_low",
-        "day_close",
-        "day_volume",
-        "day_return",
-    }
-    cols = []
-    for col in frame.columns:
-        if col in excluded:
-            continue
-        if pd.api.types.is_numeric_dtype(frame[col]) and frame[col].notna().mean() >= 0.40:
-            if frame[col].nunique(dropna=True) > 1:
-                cols.append(col)
-    return cols
-
-
-def best_threshold(y_true: np.ndarray, prob_up: np.ndarray) -> tuple[float, float]:
-    thresholds = np.linspace(0.35, 0.65, 301)
-    scores = ((prob_up[:, None] >= thresholds[None, :]) == y_true[:, None]).mean(axis=0)
-    idx = int(np.argmax(scores))
-    return float(thresholds[idx]), float(scores[idx])
-
-
-def score_auc(y_true: np.ndarray, prob_up: np.ndarray) -> float:
-    if len(np.unique(y_true)) < 2:
-        return float("nan")
-    return float(roc_auc_score(y_true, prob_up))
-
-
-def parse_args() -> argparse.Namespace:
-    parser = argparse.ArgumentParser(description="Retrain the compact NIFTY opening-direction model from Parquet data.")
-    parser.add_argument("--train-end", default=DEFAULT_TRAIN_END.date().isoformat())
-    parser.add_argument("--valid-end", default=DEFAULT_VALID_END.date().isoformat())
-    return parser.parse_args()
-
-
-def main() -> None:
-    args = parse_args()
-    train_end = pd.Timestamp(args.train_end)
-    valid_end = pd.Timestamp(args.valid_end)
-    frame = pd.read_parquet(OPENING_DATASET_PATH)
-    frame["date"] = pd.to_datetime(frame["date"], errors="coerce")
-    model_frame = frame.dropna(subset=["target"]).sort_values("date").reset_index(drop=True)
-    features = feature_columns(model_frame)
-    train_df = model_frame[model_frame["date"] <= train_end]
-    valid_df = model_frame[(model_frame["date"] > train_end) & (model_frame["date"] <= valid_end)]
-    test_df = model_frame[model_frame["date"] > valid_end]
-    if train_df.empty or valid_df.empty or test_df.empty:
-        raise RuntimeError("Training, validation, and test windows must all contain rows.")
-
-    x_train = train_df[features]
-    y_train = train_df["target"].to_numpy(dtype="int64")
-    x_valid = valid_df[features]
-    y_valid = valid_df["target"].to_numpy(dtype="int64")
-    x_test = test_df[features]
-    y_test = test_df["target"].to_numpy(dtype="int64")
-
-    extra_trees = make_pipeline(
-        SimpleImputer(strategy="median"),
-        ExtraTreesClassifier(
-            n_estimators=800,
-            max_depth=4,
-            min_samples_leaf=28,
-            max_features=0.60,
-            class_weight="balanced_subsample",
-            random_state=RANDOM_SEED + 13,
-            n_jobs=-1,
-        ),
-    )
-    logit = make_pipeline(
-        SimpleImputer(strategy="median"),
-        StandardScaler(),
-        LogisticRegression(C=0.25, class_weight="balanced", max_iter=2000, random_state=RANDOM_SEED),
-    )
-    extra_trees.fit(x_train, y_train)
-    logit.fit(x_train, y_train)
-    model = ProbabilityBlend([extra_trees, logit], np.array([0.75, 0.25]))
-    valid_prob = predict_proba_up(model, x_valid)
-    test_prob = predict_proba_up(model, x_test)
-    threshold, _ = best_threshold(y_valid, valid_prob)
-    valid_pred = apply_decision_overlays((valid_prob >= threshold).astype("int64"), valid_df, DECISION_OVERLAYS)
-    test_pred = apply_decision_overlays((test_prob >= threshold).astype("int64"), test_df, DECISION_OVERLAYS)
-    latest = frame.iloc[[-1]].copy()
-    latest_prob = predict_proba_up(model, latest[features])
-    latest_pred = apply_decision_overlays((latest_prob >= threshold).astype("int64"), latest, DECISION_OVERLAYS)
-    latest_conf = directional_confidence(latest_prob, latest_pred, threshold)
-
-    payload = {
-        "model": model,
-        "features": features,
-        "threshold": threshold,
-        "target": "same-day NIFTY 50 close > same-day NIFTY 50 open after first five 1-minute bars",
-        "model_name": "compact_extra_trees_logit_overlay",
-        "decision_overlays": DECISION_OVERLAYS,
-    }
-    joblib.dump(payload, MODEL_PATH)
-
-    summary = RetrainSummary(
-        model_name=payload["model_name"],
-        threshold=float(threshold),
-        train_rows=int(len(train_df)),
-        valid_rows=int(len(valid_df)),
-        test_rows=int(len(test_df)),
-        validation_accuracy=float(accuracy_score(y_valid, valid_pred)),
-        test_accuracy=float(accuracy_score(y_test, test_pred)),
-        validation_auc=score_auc(y_valid, valid_prob),
-        test_auc=score_auc(y_test, test_prob),
-        test_brier=float(brier_score_loss(y_test, np.clip(test_prob, 1e-6, 1 - 1e-6))),
-        latest_prediction="UP" if int(latest_pred[0]) == 1 else "DOWN",
-        latest_prob_up=float(latest_prob[0]),
-        latest_confidence=float(latest_conf[0]),
-        feature_count=int(len(features)),
-    )
-    (MODEL_DIR / "summary.json").write_text(json.dumps(asdict(summary), indent=2), encoding="utf-8")
-    pd.DataFrame([asdict(summary)]).to_csv(MODEL_DIR / "retrain_summary.csv", index=False)
-    print(asdict(summary))
-
-
-if __name__ == "__main__":
-    main()

backend/scripts/run_ist_scheduler.py

@@ -1,98 +0,0 @@
-from __future__ import annotations
-
-import sys
-import time
-from datetime import date, datetime, time as dt_time
-from pathlib import Path
-from zoneinfo import ZoneInfo
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[1]))
-from nifty_backend.runtime import (
-    CLOSE_REFRESH_READY,
-    STALE_CHECK_INTERVAL_SECONDS,
-    is_trading_day,
-    latest_saved_prediction,
-    close_refresh_due,
-    refresh_market_close_data,
-    refresh_daily_data,
-    refresh_first5_prediction,
-    refresh_stale_data_once,
-    seconds_until_next_ist_run,
-)
-
-
-IST = ZoneInfo("Asia/Kolkata")
-FIRST5_READY = dt_time(9, 20)
-
-
-def latest_prediction_date() -> date | None:
-    try:
-        raw = latest_saved_prediction().get("input_date")
-        return date.fromisoformat(str(raw)) if raw else None
-    except Exception:
-        return None
-
-
-def refresh_if_current_session_is_ready() -> None:
-    now = datetime.now(IST)
-    if not is_trading_day(now.date()) or now.time() < FIRST5_READY:
-        return
-    if latest_prediction_date() == now.date():
-        return
-    prediction = refresh_first5_prediction()
-    print(f"[scheduler] first5 prediction refreshed: {prediction.to_dict()}")
-    info = refresh_daily_data()
-    print(f"[scheduler] daily data refreshed: {info}")
-
-
-def refresh_close_data_if_due() -> None:
-    if not close_refresh_due():
-        return
-    info = refresh_market_close_data()
-    print(f"[scheduler] close data refreshed: {info}")
-
-
-def refresh_stale_data_if_due() -> None:
-    info = refresh_stale_data_once()
-    if info.get("status") == "refreshed":
-        print(f"[scheduler] stale data refreshed: {info}")
-
-
-def main() -> None:
-    print("[scheduler] NIFTY first-five-minute scheduler started.")
-    print("[scheduler] Runs the opening prediction after 09:20 IST so the 09:15-09:19 candles are complete.")
-    while True:
-        try:
-            refresh_if_current_session_is_ready()
-            refresh_close_data_if_due()
-            refresh_stale_data_if_due()
-        except Exception as exc:
-            print(f"[scheduler] current-session refresh failed: {exc}")
-        next_first5 = seconds_until_next_ist_run()
-        if next_first5 > STALE_CHECK_INTERVAL_SECONDS:
-            time.sleep(STALE_CHECK_INTERVAL_SECONDS)
-            continue
-        time.sleep(next_first5)
-        try:
-            prediction = refresh_first5_prediction()
-            print(f"[scheduler] first5 prediction refreshed: {prediction.to_dict()}")
-        except Exception as exc:
-            print(f"[scheduler] first5 refresh failed: {exc}")
-        try:
-            info = refresh_daily_data()
-            print(f"[scheduler] daily data refreshed: {info}")
-        except Exception as exc:
-            print(f"[scheduler] daily refresh failed: {exc}")
-        next_close = seconds_until_next_ist_run(CLOSE_REFRESH_READY)
-        if next_close > STALE_CHECK_INTERVAL_SECONDS:
-            time.sleep(STALE_CHECK_INTERVAL_SECONDS)
-            continue
-        time.sleep(next_close)
-        try:
-            refresh_close_data_if_due()
-        except Exception as exc:
-            print(f"[scheduler] close refresh failed: {exc}")
-
-
-if __name__ == "__main__":
-    main()