Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,4 +1,5 @@
|
|
| 1 |
import os
|
|
|
|
| 2 |
import numpy as np
|
| 3 |
import pandas as pd
|
| 4 |
import gradio as gr
|
|
@@ -6,7 +7,7 @@ import xgboost as xgb
|
|
| 6 |
from groq import Groq
|
| 7 |
|
| 8 |
# ==========================================
|
| 9 |
-
# 🔌 1. GROQ CLIENT
|
| 10 |
# ==========================================
|
| 11 |
api_key = os.environ.get("GROQ_API_KEY")
|
| 12 |
client = None
|
|
@@ -14,28 +15,47 @@ if api_key:
|
|
| 14 |
client = Groq(api_key=api_key)
|
| 15 |
|
| 16 |
# ==========================================
|
| 17 |
-
#
|
| 18 |
# ==========================================
|
| 19 |
-
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
|
| 31 |
-
|
| 32 |
-
|
| 33 |
-
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
|
| 40 |
|
| 41 |
def compute_rsi(series, period=14):
|
|
@@ -45,38 +65,69 @@ def compute_rsi(series, period=14):
|
|
| 45 |
rs = gain / loss
|
| 46 |
return 100 - (100 / (1 + rs))
|
| 47 |
|
|
|
|
| 48 |
# ==========================================
|
| 49 |
-
# 🤖 3. XGBOOST MODEL EĞİTİMİ
|
| 50 |
# ==========================================
|
| 51 |
-
def
|
| 52 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 53 |
X = df[features]
|
| 54 |
y = df["target"]
|
| 55 |
|
| 56 |
model = xgb.XGBClassifier(
|
| 57 |
-
n_estimators=50,
|
| 58 |
-
|
| 59 |
-
learning_rate=0.1,
|
| 60 |
-
objective="binary:logistic",
|
| 61 |
-
eval_metric="logloss",
|
| 62 |
-
random_state=42
|
| 63 |
)
|
| 64 |
model.fit(X, y)
|
| 65 |
return model, features
|
| 66 |
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
xgb_model, feature_cols =
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 70 |
|
| 71 |
# ==========================================
|
| 72 |
-
# 🧠 4.
|
| 73 |
# ==========================================
|
| 74 |
-
def predict_and_analyze(
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 80 |
prob_up = float(xgb_model.predict_proba(features)[0][1])
|
| 81 |
prob_down = 1 - prob_up
|
| 82 |
|
|
@@ -85,19 +136,23 @@ def predict_and_analyze(current_price, rsi_value, ma5_value, ma20_value, macd_va
|
|
| 85 |
"Düşüş (Short) Olasılığı": prob_down,
|
| 86 |
}
|
| 87 |
|
| 88 |
-
# Yön tahmini
|
| 89 |
direction = "YUKARI 📈" if prob_up > 0.5 else "AŞAĞI 📉"
|
| 90 |
confidence = max(prob_up, prob_down) * 100
|
| 91 |
|
|
|
|
| 92 |
prompt = f"""
|
| 93 |
Sen uzman bir kripto para analisti ve teknik analiz uzmanısın.
|
| 94 |
|
| 95 |
-
|
| 96 |
-
-
|
| 97 |
-
-
|
| 98 |
-
-
|
| 99 |
-
-
|
| 100 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 101 |
|
| 102 |
🤖 XGBoost Model Sonucu:
|
| 103 |
- Yükseliş Olasılığı: %{prob_up*100:.1f}
|
|
@@ -105,11 +160,11 @@ def predict_and_analyze(current_price, rsi_value, ma5_value, ma20_value, macd_va
|
|
| 105 |
- Tahmin Yönü: {direction} (Güven: %{confidence:.1f})
|
| 106 |
|
| 107 |
Türkçe bir teknik analiz raporu hazırla. Şu başlıklar bulunsun:
|
| 108 |
-
1. 🎯 Model Yorumu (XGBoost
|
| 109 |
-
2. 📈 Teknik Analiz (RSI, MA, MACD
|
| 110 |
-
3. 💡 İşlem Stratejisi (Long/Short
|
| 111 |
|
| 112 |
-
|
| 113 |
"""
|
| 114 |
|
| 115 |
if client:
|
|
@@ -122,12 +177,14 @@ def predict_and_analyze(current_price, rsi_value, ma5_value, ma20_value, macd_va
|
|
| 122 |
)
|
| 123 |
analysis_result = response.choices[0].message.content
|
| 124 |
except Exception as e:
|
| 125 |
-
analysis_result = f"Groq API
|
| 126 |
else:
|
| 127 |
-
analysis_result =
|
| 128 |
-
|
|
|
|
|
|
|
|
|
|
| 129 |
|
| 130 |
-
return analysis_result, prob_dict
|
| 131 |
|
| 132 |
# ==========================================
|
| 133 |
# 🎨 5. GRADIO ARAYÜZÜ
|
|
@@ -135,43 +192,34 @@ def predict_and_analyze(current_price, rsi_value, ma5_value, ma20_value, macd_va
|
|
| 135 |
with gr.Blocks(theme=gr.themes.Soft(primary_hue="orange", secondary_hue="slate")) as demo:
|
| 136 |
gr.Markdown(
|
| 137 |
"""
|
| 138 |
-
# ₿ BTC/USD XGBoost & Groq AI Tahmin Radarı
|
| 139 |
-
|
| 140 |
"""
|
| 141 |
)
|
| 142 |
|
| 143 |
with gr.Row():
|
| 144 |
with gr.Column():
|
| 145 |
-
gr.
|
| 146 |
-
|
| 147 |
-
|
| 148 |
-
|
| 149 |
-
|
| 150 |
-
|
| 151 |
-
with gr.Row():
|
| 152 |
-
ma5_input = gr.Number(label="MA5 ($)", value=64800.0)
|
| 153 |
-
ma20_input = gr.Number(label="MA20 ($)", value=64000.0)
|
| 154 |
-
macd_input = gr.Number(label="MACD Değeri", value=150.0)
|
| 155 |
-
|
| 156 |
-
submit_btn = gr.Button("Analiz Et ve Tahmin Üret", variant="primary")
|
| 157 |
|
| 158 |
with gr.Column():
|
| 159 |
-
gr.
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
gr.Markdown("### 🧠 Groq Yapay Zeka Analiz Raporu")
|
| 163 |
-
analysis_output = gr.Markdown(value="Verileri girdikten sonra butona basarak analizi başlatabilirsiniz.")
|
| 164 |
|
| 165 |
-
|
| 166 |
fn=predict_and_analyze,
|
| 167 |
-
inputs=[
|
| 168 |
outputs=[analysis_output, prob_output]
|
| 169 |
)
|
| 170 |
|
| 171 |
gr.Markdown(
|
| 172 |
"""
|
| 173 |
---
|
| 174 |
-
⚠️ **Yatırım Tavsiyesi Değildir**: Bu uygulama
|
| 175 |
"""
|
| 176 |
)
|
| 177 |
|
|
|
|
| 1 |
import os
|
| 2 |
+
import requests
|
| 3 |
import numpy as np
|
| 4 |
import pandas as pd
|
| 5 |
import gradio as gr
|
|
|
|
| 7 |
from groq import Groq
|
| 8 |
|
| 9 |
# ==========================================
|
| 10 |
+
# 🔌 1. GROQ CLIENT
|
| 11 |
# ==========================================
|
| 12 |
api_key = os.environ.get("GROQ_API_KEY")
|
| 13 |
client = None
|
|
|
|
| 15 |
client = Groq(api_key=api_key)
|
| 16 |
|
| 17 |
# ==========================================
|
| 18 |
+
# 📡 2. OKX PUBLIC API VERİ ÇEKME
|
| 19 |
# ==========================================
|
| 20 |
+
OKX_BASE = "https://www.okx.com/api/v5"
|
| 21 |
+
|
| 22 |
+
def fetch_okx_ticker(inst_id="BTC-USDT"):
|
| 23 |
+
"""OKX'ten anlık fiyat ve 24s verileri çeker (API Key gerekmez)."""
|
| 24 |
+
try:
|
| 25 |
+
r = requests.get(f"{OKX_BASE}/market/ticker", params={"instId": inst_id}, timeout=10)
|
| 26 |
+
r.raise_for_status()
|
| 27 |
+
d = r.json()["data"][0]
|
| 28 |
+
return {
|
| 29 |
+
"last": float(d["last"]),
|
| 30 |
+
"open24h": float(d["open24h"]),
|
| 31 |
+
"high24h": float(d["high24h"]),
|
| 32 |
+
"low24h": float(d["low24h"]),
|
| 33 |
+
"vol24h": float(d["vol24h"]),
|
| 34 |
+
"change_pct": (float(d["last"]) / float(d["open24h"]) - 1) * 100
|
| 35 |
+
}
|
| 36 |
+
except Exception as e:
|
| 37 |
+
print(f"OKX ticker error: {e}")
|
| 38 |
+
return None
|
| 39 |
+
|
| 40 |
+
|
| 41 |
+
def fetch_okx_candles(inst_id="BTC-USDT", bar="1H", limit=100):
|
| 42 |
+
"""OKX'ten mum verisi çeker ve DataFrame döndürür."""
|
| 43 |
+
try:
|
| 44 |
+
r = requests.get(f"{OKX_BASE}/market/candles",
|
| 45 |
+
params={"instId": inst_id, "bar": bar, "limit": limit},
|
| 46 |
+
timeout=10)
|
| 47 |
+
r.raise_for_status()
|
| 48 |
+
raw = r.json()["data"]
|
| 49 |
+
# OKX mum formatı: [ts, open, high, low, close, vol, ...]
|
| 50 |
+
df = pd.DataFrame(raw, columns=["ts", "open", "high", "low", "close", "vol"])
|
| 51 |
+
for col in ["open", "high", "low", "close", "vol"]:
|
| 52 |
+
df[col] = df[col].astype(float)
|
| 53 |
+
df["ts"] = pd.to_datetime(df["ts"].astype(int), unit="ms")
|
| 54 |
+
df = df.sort_values("ts").reset_index(drop=True)
|
| 55 |
+
return df
|
| 56 |
+
except Exception as e:
|
| 57 |
+
print(f"OKX candles error: {e}")
|
| 58 |
+
return None
|
| 59 |
|
| 60 |
|
| 61 |
def compute_rsi(series, period=14):
|
|
|
|
| 65 |
rs = gain / loss
|
| 66 |
return 100 - (100 / (1 + rs))
|
| 67 |
|
| 68 |
+
|
| 69 |
# ==========================================
|
| 70 |
+
# 🤖 3. XGBOOST MODEL EĞİTİMİ (OKX verisiyle)
|
| 71 |
# ==========================================
|
| 72 |
+
def train_model_from_okx():
|
| 73 |
+
"""OKX'ten çekilen mum verisiyle XGBoost modelini eğitir."""
|
| 74 |
+
df = fetch_okx_candles(bar="1H", limit=500)
|
| 75 |
+
if df is None or len(df) < 50:
|
| 76 |
+
return None, None
|
| 77 |
+
|
| 78 |
+
df["ma_5"] = df["close"].rolling(5).mean()
|
| 79 |
+
df["ma_20"] = df["close"].rolling(20).mean()
|
| 80 |
+
df["rsi"] = compute_rsi(df["close"], 14)
|
| 81 |
+
df["macd"] = df["close"].ewm(span=12).mean() - df["close"].ewm(span=26).mean()
|
| 82 |
+
df["target"] = (df["close"].shift(-1) > df["close"]).astype(int)
|
| 83 |
+
|
| 84 |
+
df = df.dropna().reset_index(drop=True)
|
| 85 |
+
features = ["ma_5", "ma_20", "rsi", "macd", "vol"]
|
| 86 |
X = df[features]
|
| 87 |
y = df["target"]
|
| 88 |
|
| 89 |
model = xgb.XGBClassifier(
|
| 90 |
+
n_estimators=50, max_depth=3, learning_rate=0.1,
|
| 91 |
+
objective="binary:logistic", eval_metric="logloss", random_state=42
|
|
|
|
|
|
|
|
|
|
|
|
|
| 92 |
)
|
| 93 |
model.fit(X, y)
|
| 94 |
return model, features
|
| 95 |
|
| 96 |
+
|
| 97 |
+
# Başlangıçta modeli OKX verisiyle eğit
|
| 98 |
+
xgb_model, feature_cols = train_model_from_okx()
|
| 99 |
+
if xgb_model is None:
|
| 100 |
+
# Fallback: OKX erişilemezse basit bir dummy model
|
| 101 |
+
xgb_model = xgb.XGBClassifier(n_estimators=10, max_depth=2)
|
| 102 |
+
dummy_X = np.random.rand(20, 5)
|
| 103 |
+
dummy_y = np.random.randint(0, 2, 20)
|
| 104 |
+
xgb_model.fit(dummy_X, dummy_y)
|
| 105 |
+
feature_cols = ["ma_5", "ma_20", "rsi", "macd", "vol"]
|
| 106 |
|
| 107 |
# ==========================================
|
| 108 |
+
# 🧠 4. ANALİZ MOTORU
|
| 109 |
# ==========================================
|
| 110 |
+
def predict_and_analyze():
|
| 111 |
+
# 1) OKX'ten canlı veri çek
|
| 112 |
+
ticker = fetch_okx_ticker()
|
| 113 |
+
candles = fetch_okx_candles(limit=30)
|
| 114 |
+
|
| 115 |
+
if ticker is None or candles is None:
|
| 116 |
+
return "❌ OKX API'sine erişilemedi. Lütfen daha sonra tekrar deneyin.", {}
|
| 117 |
+
|
| 118 |
+
last_price = ticker["last"]
|
| 119 |
+
change_pct = ticker["change_pct"]
|
| 120 |
+
vol24h = ticker["vol24h"]
|
| 121 |
+
|
| 122 |
+
# 2) Teknik göstergeler
|
| 123 |
+
close = candles["close"]
|
| 124 |
+
ma5 = close.rolling(5).mean().iloc[-1]
|
| 125 |
+
ma20 = close.rolling(20).mean().iloc[-1]
|
| 126 |
+
rsi = compute_rsi(close, 14).iloc[-1]
|
| 127 |
+
macd = (close.ewm(span=12).mean() - close.ewm(span=26).mean()).iloc[-1]
|
| 128 |
+
|
| 129 |
+
# 3) XGBoost tahmini
|
| 130 |
+
features = np.array([[ma5, ma20, rsi, macd, vol24h]])
|
| 131 |
prob_up = float(xgb_model.predict_proba(features)[0][1])
|
| 132 |
prob_down = 1 - prob_up
|
| 133 |
|
|
|
|
| 136 |
"Düşüş (Short) Olasılığı": prob_down,
|
| 137 |
}
|
| 138 |
|
|
|
|
| 139 |
direction = "YUKARI 📈" if prob_up > 0.5 else "AŞAĞI 📉"
|
| 140 |
confidence = max(prob_up, prob_down) * 100
|
| 141 |
|
| 142 |
+
# 4) Groq analizi
|
| 143 |
prompt = f"""
|
| 144 |
Sen uzman bir kripto para analisti ve teknik analiz uzmanısın.
|
| 145 |
|
| 146 |
+
📡 OKX Canlı Verileri (BTC/USDT):
|
| 147 |
+
- Son Fiyat: ${last_price:,.2f}
|
| 148 |
+
- 24s Değişim: %{change_pct:+.2f}
|
| 149 |
+
- 24s Hacim: {vol24h:,.0f} BTC
|
| 150 |
+
- 24s Yüksek/Düşük: ${ticker['high24h']:,.2f} / ${ticker['low24h']:,.2f}
|
| 151 |
+
|
| 152 |
+
📊 Teknik Göstergeler (1H):
|
| 153 |
+
- RSI (14): {rsi:.2f}
|
| 154 |
+
- MA5: ${ma5:,.2f} | MA20: ${ma20:,.2f}
|
| 155 |
+
- MACD: {macd:.4f}
|
| 156 |
|
| 157 |
🤖 XGBoost Model Sonucu:
|
| 158 |
- Yükseliş Olasılığı: %{prob_up*100:.1f}
|
|
|
|
| 160 |
- Tahmin Yönü: {direction} (Güven: %{confidence:.1f})
|
| 161 |
|
| 162 |
Türkçe bir teknik analiz raporu hazırla. Şu başlıklar bulunsun:
|
| 163 |
+
1. 🎯 Model Yorumu (XGBoost + teknik göstergelerin mantığı)
|
| 164 |
+
2. 📈 Teknik Analiz (RSI, MA, MACD; destek/direnç)
|
| 165 |
+
3. 💡 İşlem Stratejisi (Long/Short, stop-loss, take-profit, risk yönetimi)
|
| 166 |
|
| 167 |
+
Yatırım tavsiyesi olmadığını vurgula. Rasyonel ve profesyonel ol.
|
| 168 |
"""
|
| 169 |
|
| 170 |
if client:
|
|
|
|
| 177 |
)
|
| 178 |
analysis_result = response.choices[0].message.content
|
| 179 |
except Exception as e:
|
| 180 |
+
analysis_result = f"Groq API Hatası: {e}"
|
| 181 |
else:
|
| 182 |
+
analysis_result = "GROQ_API_KEY bulunamadı! Hugging Face Secrets'a ekleyin."
|
| 183 |
+
|
| 184 |
+
# Başlık bilgisi
|
| 185 |
+
header = f"**OKX Canlı: ${last_price:,.2f}** (%{change_pct:+.2f}) | RSI: {rsi:.1f} | MA5: ${ma5:,.0f} | MA20: ${ma20:,.0f}\n\n---\n\n"
|
| 186 |
+
return header + analysis_result, prob_dict
|
| 187 |
|
|
|
|
| 188 |
|
| 189 |
# ==========================================
|
| 190 |
# 🎨 5. GRADIO ARAYÜZÜ
|
|
|
|
| 192 |
with gr.Blocks(theme=gr.themes.Soft(primary_hue="orange", secondary_hue="slate")) as demo:
|
| 193 |
gr.Markdown(
|
| 194 |
"""
|
| 195 |
+
# ₿ BTC/USD XGBoost & Groq AI Tahmin Radarı (OKX Canlı)
|
| 196 |
+
Fiyat, hacim ve teknik göstergeler **OKX halka açık API**'sinden otomatik çekilir. **XGBoost** yön tahmini yapar, **Groq** analiz raporu üretir.
|
| 197 |
"""
|
| 198 |
)
|
| 199 |
|
| 200 |
with gr.Row():
|
| 201 |
with gr.Column():
|
| 202 |
+
analyze_btn = gr.Button("🔄 OKX'ten Veri Çek ve Analiz Et", variant="primary", size="lg")
|
| 203 |
+
gr.Markdown(
|
| 204 |
+
"""
|
| 205 |
+
**Not**: OKX API'si her 2 saniyede 20 istek sınırına sahiptir [citation:7]. Butona çok hızlı basmayın.
|
| 206 |
+
"""
|
| 207 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 208 |
|
| 209 |
with gr.Column():
|
| 210 |
+
prob_output = gr.Label(label="XGBoost Yön Tahmini")
|
| 211 |
+
analysis_output = gr.Markdown(value="Butona basarak OKX'ten canlı veri çekip analizi başlatın.")
|
|
|
|
|
|
|
|
|
|
| 212 |
|
| 213 |
+
analyze_btn.click(
|
| 214 |
fn=predict_and_analyze,
|
| 215 |
+
inputs=[],
|
| 216 |
outputs=[analysis_output, prob_output]
|
| 217 |
)
|
| 218 |
|
| 219 |
gr.Markdown(
|
| 220 |
"""
|
| 221 |
---
|
| 222 |
+
⚠️ **Yatırım Tavsiyesi Değildir**: Bu uygulama eğitim amaçlıdır. Kripto paralar yüksek risk içerir.
|
| 223 |
"""
|
| 224 |
)
|
| 225 |
|