Update app.py

app.py

@@ -1,4 +1,8 @@
-# app.py
+# ===============================================================
+#  Forex EMA + Dynamic Normalization API (Model B)
+#  Versi Final – Aman untuk Hugging Face Spaces (tanpa cache)
+# ===============================================================
+
 from fastapi import FastAPI
 from pydantic import BaseModel
 import pandas as pd
@@ -6,149 +10,122 @@ import numpy as np
 import yfinance as yf
 from datetime import datetime, timedelta
 import logging
+import tempfile
 import os
 
-# -----------------------
-# Logging
-# -----------------------
-logging.basicConfig(level=logging.INFO, format="%(asctime)s — %(levelname)s — %(message)s")
-logger = logging.getLogger("modelb_api")
+# ===============================================================
+# Konfigurasi Logging
+# ===============================================================
+logging.basicConfig(
+    level=logging.INFO,
+    format="%(asctime)s — %(levelname)s — %(message)s"
+)
+logger = logging.getLogger(__name__)
 
-# -----------------------
-# App & config
-# -----------------------
+# ===============================================================
+# Konfigurasi FastAPI
+# ===============================================================
 app = FastAPI(
     title="Model B – EMA & Dynamic Scaling API",
     description="API untuk menghitung EMA, normalisasi, dan analisis tren otomatis berdasarkan data yfinance",
-    version="2.1"
+    version="2.3"
 )
 
 PAIR = "EURUSD=X"
-BASE_WINDOW = 60  # jumlah hari historical yang diambil untuk jaga-jaga agar EMA50 bisa dihitung
+BASE_WINDOW = 60
 
-# -----------------------
-# Request schema
-# -----------------------
+# Matikan cache yfinance agar tidak menulis ke /.cache
+os.environ["YFINANCE_CACHE_DISABLE"] = "1"
+os.environ["YFINANCE_NO_TZ_CACHE"] = "1"
+yf.set_tz_cache_location(tempfile.gettempdir())
+
+# ===============================================================
+# Schema
+# ===============================================================
 class DateRange(BaseModel):
-    start_date: str  # ISO date 'YYYY-MM-DD'
-    end_date: str    # ISO date 'YYYY-MM-DD'
+    start_date: str
+    end_date: str
+
 
-def clean_yf_csv(df_raw):
+# ===============================================================
+# Helper Function – Load data langsung dari yfinance
+# ===============================================================
+def load_yf_data(pair, start, end):
     """
-    Membersihkan struktur CSV seperti hasil export yfinance:
-    - baris pertama: nama kolom
-    - baris kedua: ticker
-    - baris ketiga: header date kosong
-    - sisakan hanya kolom ['date', 'close']
+    Ambil data yfinance tanpa cache, hanya return kolom ['date', 'close'].
     """
-    # Jika baris pertama bukan data (mengandung "Price,Close,High,..."), kita re-parse manual
-    if "Price" in df_raw.columns[0]:
-        logger.info("Detected raw yfinance export format, cleaning...")
-        # Baca ulang file tanpa header dulu
-        df_raw.columns = [c.strip().lower() for c in df_raw.columns]
-        # Hapus baris ke-2 dan ke-3
-        df_raw = df_raw.drop(index=[0, 1])
-        # Ganti kolom 'price' menjadi 'date'
-        df_raw = df_raw.rename(columns={"price": "date"})
-        # Pastikan hanya ambil kolom 'date' dan 'close'
-        keep_cols = [c for c in df_raw.columns if c in ["date", "close"]]
-        df_raw = df_raw[keep_cols]
-        # Ubah tipe data tanggal
-        df_raw["date"] = pd.to_datetime(df_raw["date"], errors="coerce")
-        df_raw = df_raw.dropna(subset=["date"])
-        df_raw["close"] = pd.to_numeric(df_raw["close"], errors="coerce")
-        df_raw = df_raw.dropna(subset=["close"])
-        df_raw = df_raw.reset_index(drop=True)
-    return df_raw
+    try:
+        df = yf.download(pair, start=start, end=end, auto_adjust=True, progress=False)
 
-def load_yf_data(pair, start, end):
-    """Download yfinance data and ensure 'date' and 'close' columns exist."""
-    df_raw = yf.download(pair, start=start, end=end, auto_adjust=True, progress=False)
-    if df_raw.empty:
-        raise ValueError("No data from yfinance for that range.")
+        if df.empty:
+            raise ValueError("YFinance gagal mengambil data, data kosong.")
 
-    # Jika data CSV seperti contoh kamu (ada baris Ticker dan Price)
-    if "Price" in df_raw.columns or "Ticker" in df_raw.columns:
-        df_raw = clean_yf_csv(df_raw)
-        return df_raw
+        # Jika MultiIndex, ambil level pertama
+        if isinstance(df.columns, pd.MultiIndex):
+            df.columns = df.columns.get_level_values(0)
 
-    # flatten MultiIndex columns
-    if isinstance(df_raw.columns, pd.MultiIndex):
-        df_raw.columns = df_raw.columns.get_level_values(0)
+        # Ambil hanya kolom Close
+        close_col = [c for c in df.columns if "close" in c.lower()]
+        if not close_col:
+            raise ValueError(f"Tidak ada kolom 'Close' ditemukan di {df.columns.tolist()}")
 
-    df = df_raw.reset_index()[["Date", "Close"]].rename(columns={"Date": "date", "Close": "close"})
-    df["date"] = pd.to_datetime(df["date"]).dt.normalize()
-    return df
+        df = df.reset_index()[["Date", close_col[0]]]
+        df.columns = ["date", "close"]
+        df["date"] = pd.to_datetime(df["date"]).dt.date
+        df["close"] = pd.to_numeric(df["close"], errors="coerce")
+        df = df.dropna(subset=["close"]).reset_index(drop=True)
 
+        logger.info(f"✅ Berhasil ambil data {len(df)} baris dari {pair}")
+        return df
 
-# -----------------------
-# Utility functions
-# -----------------------
+    except Exception as e:
+        logger.error(f"❌ Error load_yf_data(): {e}", exc_info=True)
+        raise ValueError(str(e))
+
+
+# ===============================================================
+# Helper Function – Manual EMA
+# ===============================================================
 def ema_manual(prices, span):
-    """
-    Manual EMA calculation:
-      - prices: iterable numeric
-      - span: int (e.g. 20 or 50)
-    Returns list of same length; first (span-1) entries are NaN, entry at index span-1 is SMA seed.
-    """
-    prices = np.asarray(prices, dtype=float)
-    n = len(prices)
-    if n == 0:
-        return [np.nan] * 0
-    if span <= 0:
-        raise ValueError("span must be > 0")
-    ema = [np.nan] * n
-    alpha = 2.0 / (span + 1.0)
-
-    # Not enough data -> return NaNs (keputusan: tetap mengembalikan NaN untuk indeks sebelum span-1)
-    if n < span:
-        logger.warning(f"Not enough data for EMA span={span} (have {n} < needed {span}), returning NaNs.")
-        return ema
-
-    # seed with SMA at index span-1
-    seed = float(np.mean(prices[:span]))
-    ema[span - 1] = seed
-
-    # recursive EMA
-    for i in range(span, n):
-        ema[i] = alpha * prices[i] + (1.0 - alpha) * ema[i - 1]
+    if len(prices) < span:
+        return [np.nan] * len(prices)
+
+    ema = [np.nan] * len(prices)
+    alpha = 2 / (span + 1)
+    ema[span - 1] = np.mean(prices[:span])
+
+    for i in range(span, len(prices)):
+        ema[i] = alpha * prices[i] + (1 - alpha) * ema[i - 1]
 
     return ema
 
-def get_dynamic_minmax(pair=PAIR, window_days=BASE_WINDOW):
-    """Download recent window and return min/max of Close to use for normalization."""
-    today = datetime.utcnow().date()
-    start = today - timedelta(days=window_days)
-    logger.info(f"Fetching recent data for dynamic min/max: {start} -> {today}")
-    try:
-        df = yf.download(pair, start=start, end=today + timedelta(days=1), auto_adjust=True, progress=False)
-    except Exception as e:
-        logger.error("yfinance download failed for dynamic minmax: %s", e, exc_info=True)
-        raise
 
-    if df is None or df.empty:
-        raise ValueError("Failed to fetch recent prices for dynamic min/max")
+# ===============================================================
+# Helper Function – Dynamic Scaling
+# ===============================================================
+def get_dynamic_minmax():
+    today = datetime.now().date()
+    start = today - timedelta(days=BASE_WINDOW)
+    df = load_yf_data(PAIR, start, today + timedelta(days=1))
+    close_min = df["close"].min()
+    close_max = df["close"].max()
+    logger.info(f"Dynamic Min/Max Close: {close_min:.5f} / {close_max:.5f}")
+    return float(close_min), float(close_max)
 
-    close_min = float(df["Close"].min())
-    close_max = float(df["Close"].max())
-    logger.info("dynamic min/max: %s / %s", close_min, close_max)
-    return close_min, close_max
 
 def normalize_close(value, close_min, close_max):
     if close_max == close_min:
         return 0.5
-    return float((value - close_min) / (close_max - close_min))
+    return (value - close_min) / (close_max - close_min)
 
-def analyze_trend(latest_row):
-    """
-    Return simple analysis dict based on EMA20 vs EMA50 and gap percent.
-    """
-    ema20 = latest_row.get("EMA20", np.nan)
-    ema50 = latest_row.get("EMA50", np.nan)
-    close = latest_row.get("close", np.nan)
 
-    if np.isnan(ema20) or np.isnan(ema50):
-        return {"trend": "unknown", "strength": "unknown", "price_position": "unknown", "ema_gap_percent": None}
+# ===============================================================
+# Helper Function – Trend Analysis
+# ===============================================================
+def analyze_trend(latest_row):
+    ema20 = latest_row["EMA20"]
+    ema50 = latest_row["EMA50"]
+    close = latest_row["close"]
 
     if ema20 > ema50:
         trend = "bullish"
@@ -157,14 +134,10 @@ def analyze_trend(latest_row):
     else:
         trend = "neutral"
 
-    if ema50 == 0 or np.isnan(ema50):
-        gap_pct = 0.0
-    else:
-        gap_pct = abs(ema20 - ema50) / abs(ema50) * 100.0
-
-    if gap_pct > 0.3:
+    diff = abs(ema20 - ema50) / ema50 * 100 if ema50 != 0 else 0
+    if diff > 0.3:
         strength = "strong"
-    elif gap_pct > 0.1:
+    elif diff > 0.1:
         strength = "moderate"
     else:
         strength = "weak"
@@ -180,139 +153,94 @@ def analyze_trend(latest_row):
         "trend": trend,
         "strength": strength,
         "price_position": price_position,
-        "ema_gap_percent": round(gap_pct, 4)
+        "ema_gap_percent": round(diff, 3)
     }
 
-# -----------------------
-# Endpoints
-# -----------------------
+
+# ===============================================================
+# Endpoint: /analyze
+# ===============================================================
 @app.post("/analyze")
 def analyze_ema_endpoint(input_data: DateRange):
-    """
-    Return time series data (date, close, EMA20, EMA50, norm_close) for plotting.
-    If requested window is too short to compute EMA50, endpoint automatically uses earlier data
-    by extending start_date backward by BASE_WINDOW days.
-    """
     try:
-        logger.info("Received /analyze request: %s -> %s", input_data.start_date, input_data.end_date)
-        start_date = pd.to_datetime(input_data.start_date).date()
-        end_date = pd.to_datetime(input_data.end_date).date()
-    except Exception as e:
-        logger.error("Invalid date format: %s", e)
-        return {"status": "error", "message": "Invalid date format. Use YYYY-MM-DD."}
+        start_date = pd.to_datetime(input_data.start_date)
+        end_date = pd.to_datetime(input_data.end_date)
+        extended_start = start_date - timedelta(days=60)
 
-    # extend start to ensure enough history (for EMA50)
-    extended_start = start_date - timedelta(days=BASE_WINDOW)
+        df = load_yf_data(PAIR, extended_start, end_date + timedelta(days=1))
+        if len(df) < 50:
+            return {"status": "error", "message": "Data terlalu sedikit (butuh minimal 50 hari)."}
 
-    try:
-        df_raw = yf.download(PAIR, start=extended_start, end=end_date + timedelta(days=1), auto_adjust=True, progress=False)
-    except Exception as e:
-        logger.error("yfinance download failed: %s", e, exc_info=True)
-        return {"status": "error", "message": f"Failed to fetch data from yfinance: {e}"}
-
-    if df_raw is None or df_raw.empty:
-        logger.warning("No data returned from yfinance for the requested range")
-        return {"status": "error", "message": "No price data for requested dates"}
-
-    df = df_raw.reset_index()[["Date", "Close"]].rename(columns={"Date": "date", "Close": "close"})
-    df["date"] = pd.to_datetime(df["date"]).dt.normalize()
-    logger.info("Downloaded rows: %d", len(df))
-
-    if len(df) < 50:
-        msg = f"Insufficient data points after extension ({len(df)}). Need at least 50 for EMA50."
-        logger.error(msg)
-        return {"status": "error", "message": msg}
-
-    # compute EMAs
-    close_list = df["close"].tolist()
-    df["EMA20"] = ema_manual(close_list, 20)
-    df["EMA50"] = ema_manual(close_list, 50)
-
-    # drop rows where EMA values are NaN (i.e., before we have enough seed)
-    df = df.dropna(subset=["EMA20", "EMA50"]).reset_index(drop=True)
-    if df.empty:
-        return {"status": "error", "message": "After computing EMAs no usable rows remain."}
+        df["EMA20"] = ema_manual(df["close"].values.tolist(), 20)
+        df["EMA50"] = ema_manual(df["close"].values.tolist(), 50)
+        df = df.dropna().reset_index(drop=True)
 
-    # dynamic min/max and normalization
-    try:
         close_min, close_max = get_dynamic_minmax()
+        df["norm_close"] = df["close"].apply(lambda x: normalize_close(x, close_min, close_max))
+
+        chart_data = {
+            "dates": [str(d) for d in df["date"].tolist()],
+            "close": df["close"].round(6).tolist(),
+            "EMA20": df["EMA20"].round(6).tolist(),
+            "EMA50": df["EMA50"].round(6).tolist(),
+            "norm_close": df["norm_close"].round(6).tolist(),
+            "min_close": close_min,
+            "max_close": close_max
+        }
+
+        return {
+            "status": "ok",
+            "pair": PAIR,
+            "start_date": str(start_date.date()),
+            "end_date": str(end_date.date()),
+            "data_points": len(df),
+            "chart_data": chart_data
+        }
+
     except Exception as e:
-        logger.warning("Dynamic min/max fetch failed: %s", e)
-        close_min, close_max = float(df["close"].min()), float(df["close"].max())
-
-    df["norm_close"] = df["close"].apply(lambda x: normalize_close(x, close_min, close_max))
-
-    chart_data = {
-        "dates": df["date"].dt.strftime("%Y-%m-%d").tolist(),
-        "close": [float(x) for x in df["close"].tolist()],
-        "ema20": [float(x) for x in df["EMA20"].tolist()],
-        "ema50": [float(x) for x in df["EMA50"].tolist()],
-        "norm_close": [float(x) for x in df["norm_close"].tolist()],
-        "min_close": float(close_min),
-        "max_close": float(close_max),
-    }
+        logger.error(f"Error di /analyze: {e}", exc_info=True)
+        return {"status": "error", "message": str(e)}
 
-    logger.info("Analyze success -> points: %d", len(df))
-    return {
-        "status": "ok",
-        "pair": PAIR,
-        "requested_start": str(start_date),
-        "requested_end": str(end_date),
-        "data_points": len(df),
-        "chart_data": chart_data
-    }
 
+# ===============================================================
+# Endpoint: /summary
+# ===============================================================
 @app.post("/summary")
 def ema_summary_endpoint(input_data: DateRange):
-    """
-    Return last available close, EMA20, EMA50 and a short trend analysis dictionary.
-    """
     try:
-        logger.info("Received /summary request: %s -> %s", input_data.start_date, input_data.end_date)
-        start_date = pd.to_datetime(input_data.start_date).date()
-        end_date = pd.to_datetime(input_data.end_date).date()
-    except Exception as e:
-        logger.error("Invalid date input for /summary: %s", e)
-        return {"status": "error", "message": "Invalid date format. Use YYYY-MM-DD."}
-
-    extended_start = start_date - timedelta(days=BASE_WINDOW)
+        start_date = pd.to_datetime(input_data.start_date)
+        end_date = pd.to_datetime(input_data.end_date)
+        extended_start = start_date - timedelta(days=60)
+
+        df = load_yf_data(PAIR, extended_start, end_date + timedelta(days=1))
+        if len(df) < 50:
+            return {"status": "error", "message": "Data terlalu sedikit (butuh minimal 50 hari)."}
+
+        df["EMA20"] = ema_manual(df["close"].values.tolist(), 20)
+        df["EMA50"] = ema_manual(df["close"].values.tolist(), 50)
+        df = df.dropna().reset_index(drop=True)
+
+        latest = df.iloc[-1]
+        analysis = analyze_trend(latest)
+
+        return {
+            "status": "ok",
+            "pair": PAIR,
+            "as_of_date": str(latest["date"]),
+            "close": round(float(latest["close"]), 6),
+            "EMA20": round(float(latest["EMA20"]), 6),
+            "EMA50": round(float(latest["EMA50"]), 6),
+            "trend_analysis": analysis
+        }
 
-    try:
-        df_raw = yf.download(PAIR, start=extended_start, end=end_date + timedelta(days=1), auto_adjust=True, progress=False)
     except Exception as e:
-        logger.error("yfinance download failed for /summary: %s", e, exc_info=True)
-        return {"status": "error", "message": f"Failed to fetch data from yfinance: {e}"}
-
-    if df_raw is None or df_raw.empty:
-        return {"status": "error", "message": "No price data for requested dates"}
-
-    df = df_raw.reset_index()[["Date", "Close"]].rename(columns={"Date": "date", "Close": "close"})
-    df["date"] = pd.to_datetime(df["date"]).dt.normalize()
+        logger.error(f"Error di /summary: {e}", exc_info=True)
+        return {"status": "error", "message": str(e)}
 
-    if len(df) < 50:
-        return {"status": "error", "message": f"Insufficient data (have {len(df)} rows). Need >=50 for EMA50."}
-
-    close_list = df["close"].tolist()
-    df["EMA20"] = ema_manual(close_list, 20)
-    df["EMA50"] = ema_manual(close_list, 50)
-    df = df.dropna(subset=["EMA20", "EMA50"]).reset_index(drop=True)
-
-    if df.empty:
-        return {"status": "error", "message": "No usable rows after EMA computation."}
-
-    latest = df.iloc[-1]
-    analysis = analyze_trend(latest)
-
-    return {
-        "status": "ok",
-        "pair": PAIR,
-        "as_of_date": latest["date"].strftime("%Y-%m-%d"),
-        "close": float(latest["close"]),
-        "ema20": float(latest["EMA20"]),
-        "ema50": float(latest["EMA50"]),
-        "trend_analysis": analysis
-    }
 
+# ===============================================================
+# Root Test
+# ===============================================================
 @app.get("/")
 def root():
     return {"message": "Model B API (EMA + Trend Summary) aktif 🚀"}