change app

app.py

@@ -8,7 +8,6 @@ import numpy as np
 import os
 import requests
 import yfinance as yf
-import pandas as pd
 
 from huggingface_hub import hf_hub_download
 from model import LSTMModel
@@ -24,7 +23,8 @@ logging.basicConfig(level=logging.INFO)
 # CONFIG
 # ===========================
 HF_MODEL_REPO = "gaidasalsaa/lstm-exchange-rate-prediction-model"
-EXCHANGE_API_KEY = os.getenv("8b579e2eb898f797a652e3ef")
+
+EXCHANGE_API_KEY = os.getenv("EXCHANGE_API_KEY")
 EXCHANGE_API_URL = "https://v6.exchangerate-api.com/v6"
 
 MODEL_MAP = {
@@ -104,7 +104,7 @@ class PredictionResponse(BaseModel):
     data: Optional[dict] = None
 
 # ===========================
-# REAL-TIME EXCHANGE RATE
+# REAL-TIME RATE
 # ===========================
 def get_realtime_rate(base, target):
     url = f"{EXCHANGE_API_URL}/{EXCHANGE_API_KEY}/pair/{base}/{target}"
@@ -112,11 +112,27 @@ def get_realtime_rate(base, target):
     data = r.json()
 
     if data.get("result") != "success":
-        logger.error(f"ExchangeRate API error: {data}")
-        raise ValueError("Failed to fetch real-time rate")
+        raise ValueError("ExchangeRate API failed")
 
     return float(data["conversion_rate"])
 
+# ===========================
+# HISTORICAL DATA (YAHOO)
+# ===========================
+def fetch_last_30_days(base, target):
+    symbol = f"{base}{target}=X"
+
+    df = yf.download(
+        symbol,
+        period="60d",
+        interval="1d",
+        progress=False
+    )
+
+    if df.empty or len(df) < LOOKBACK:
+        return None
+
+    return df["Close"].dropna().tail(LOOKBACK).values.reshape(-1, 1)
 
 # ===========================
 # INFERENCE
@@ -131,16 +147,11 @@ def predict_forex(base, target, horizon):
     if model is None or scaler is None:
         return None, None
 
-    # 1. Ambil harga real-time
-    current_price = get_realtime_rate(base, target)
-
-    # 2. Buat input sequence (LOOKBACK)
-    # dummy historical values (stabil & aman)
-    last_values = np.full((LOOKBACK - 1, 1), current_price)
-    prices = np.vstack([last_values, [[current_price]]])
+    prices = fetch_last_30_days(base, target)
+    if prices is None:
+        return None, None
 
     scaled = scaler.transform(prices)
-
     X = create_input_sequence(scaled, LOOKBACK)
     X = torch.tensor(X, dtype=torch.float32)
 
@@ -148,24 +159,7 @@ def predict_forex(base, target, horizon):
         preds = model(X).numpy()
 
     preds = scaler.inverse_transform(preds.reshape(-1, 1)).flatten()
-    return current_price, preds.tolist()
-
-# Mengambil 30 hari terakhir dari yfinance
-def fetch_last_30_days(base, target):
-    symbol = f"{base}{target}=X"
-
-    df = yf.download(
-        symbol,
-        period="60d",   # ambil lebih untuk jaga-jaga
-        interval="1d",
-        progress=False
-    )
-
-    if df.empty or len(df) < 30:
-        return None
-
-    prices = df["Close"].dropna().tail(30).values.tolist()
-    return prices
+    return prices[-1][0], preds.tolist()
 
 # ===========================
 # ROUTE
@@ -173,36 +167,25 @@ def fetch_last_30_days(base, target):
 @app.post("/predict", response_model=PredictionResponse)
 def predict(req: PredictionRequest):
 
-    prices = req.recent_prices
-
-    # 🔥 Jika frontend tidak kirim data → backend ambil sendiri
-    if not prices or len(prices) < LOOKBACK:
-        logger.info("Fetching prices from Yahoo Finance...")
-        prices = fetch_last_30_days(
-            req.base_currency,
-            req.target_currency
-        )
+    current_price, preds = predict_forex(
+        req.base_currency,
+        req.target_currency,
+        req.horizon
+    )
 
-    if prices is None:
+    if preds is None:
         return PredictionResponse(
-            message="Failed to fetch historical prices",
+            message="Prediction failed",
             data=None
         )
 
-    preds = predict_forex(
-        req.base_currency,
-        req.target_currency,
-        req.horizon,
-        prices
-    )
-
     return PredictionResponse(
         message="Prediction success",
         data={
             "base": req.base_currency,
             "target": req.target_currency,
             "horizon": req.horizon,
-            "last_price": prices[-1],
+            "current_price": current_price,
             "predictions": preds
         }
     )