Add application file

app.py

@@ -0,0 +1,244 @@
+import gradio as gr
+import pandas as pd
+import plotly.express as px
+import re
+import io
+import os
+
+# ==========================================
+# 1. 数据读取引擎 (防弹版 - 保持不变)
+# ==========================================
+
+def clean_and_read_file(file_path):
+    """
+    Robust file reader:
+    1. Handles .xlsx masquerading as .csv
+    2. Cleans garbage tags like 
+    3. Fixes broken lines
+    """
+    if not file_path or not os.path.exists(file_path):
+        return pd.DataFrame()
+
+    # --- Strategy A: Try reading as Excel ---
+    try:
+        df = pd.read_excel(file_path)
+        return df
+    except:
+        pass 
+
+    # --- Strategy B: Read as Text ---
+    raw_data = b""
+    try:
+        with open(file_path, 'rb') as f:
+            raw_data = f.read()
+    except Exception as e:
+        print(f"File read error: {e}")
+        return pd.DataFrame()
+
+    # Decode
+    content = ""
+    for enc in ['utf-8', 'gb18030', 'gbk']:
+        try:
+            content = raw_data.decode(enc)
+            break
+        except UnicodeDecodeError:
+            continue
+    if not content:
+        content = raw_data.decode('utf-8', errors='replace')
+
+    # --- Cleaning ---
+    content = re.sub(r"\\", "", content)
+    
+    lines = content.splitlines()
+    cleaned_lines = []
+    buffer = ""
+    date_pattern = re.compile(r'^\s*202\d-\d{2}-\d{2}') 
+    
+    for line in lines:
+        line = line.strip()
+        if not line: continue
+        
+        is_header = "Date" in line and ("," in line)
+        is_date_row = date_pattern.match(line) is not None
+        
+        if is_header or is_date_row:
+            if buffer: cleaned_lines.append(buffer)
+            buffer = line
+        else:
+            buffer += " " + line
+            
+    if buffer: cleaned_lines.append(buffer)
+    
+    csv_content = "\n".join(cleaned_lines)
+    try:
+        df = pd.read_csv(io.StringIO(csv_content))
+    except:
+        try:
+            df = pd.read_csv(io.StringIO(csv_content), sep=None, engine='python')
+        except:
+            return pd.DataFrame()
+
+    return df
+
+# ==========================================
+# 2. 数据处理
+# ==========================================
+
+def process_gpu_data(df):
+    if df.empty: return df
+    df.columns = [str(c).strip() for c in df.columns]
+
+    if 'Date' in df.columns:
+        df['Date'] = pd.to_datetime(df['Date'], errors='coerce')
+    
+    def clean_currency(x):
+        if isinstance(x, (int, float)): return float(x)
+        if isinstance(x, str):
+            match = re.search(r'(\d+\.?\d*)', x)
+            return float(match.group(1)) if match else 0.0
+        return 0.0
+
+    target_col = None
+    if 'Cloud Rent (/hr)' in df.columns:
+        target_col = 'Cloud Rent (/hr)'
+    else:
+        for c in df.columns:
+            if 'Rent' in c or '/hr' in c:
+                target_col = c
+                break
+    
+    if target_col:
+        df['Rent_Price_Num'] = df[target_col].apply(clean_currency)
+    
+    return df
+
+def process_llm_data(df):
+    if df.empty: return df
+    df.columns = [str(c).strip() for c in df.columns]
+    
+    if 'Date' in df.columns:
+        df['Date'] = pd.to_datetime(df['Date'], errors='coerce')
+        
+    return df
+
+# ==========================================
+# 3. 绘图逻辑
+# ==========================================
+
+def plot_gpu_trends(df):
+    if df is None or df.empty or 'Rent_Price_Num' not in df.columns:
+        return None
+    
+    plot_df = df.dropna(subset=['Date', 'Rent_Price_Num'])
+    if plot_df.empty: return None
+
+    chip_col = 'Chip' if 'Chip' in df.columns else df.columns[1]
+
+    fig = px.line(
+        plot_df, 
+        x='Date', 
+        y='Rent_Price_Num', 
+        color=chip_col if chip_col in df.columns else None,
+        title='GPU Cloud Rental Price Trends ($/hr)',
+        labels={'Rent_Price_Num': 'Price ($/hr)', 'Date': 'Date'},
+        markers=True
+    )
+    return fig
+
+def plot_llm_trends(df):
+    """绘制所有列的趋势，不再需要 selection"""
+    if df is None or df.empty:
+        return None
+        
+    # 自动选择除了 Date 以外的所有列
+    value_vars = [c for c in df.columns if c != 'Date']
+    if not value_vars:
+        return None
+
+    plot_df = df[['Date'] + value_vars].copy().dropna(subset=['Date'])
+    
+    # Melt
+    df_long = plot_df.melt(id_vars=['Date'], var_name='Model', value_name='Price')
+    
+    fig = px.line(
+        df_long,
+        x='Date',
+        y='Price',
+        color='Model',
+        title='LLM API Price Trends',
+        labels={'Price': 'Price', 'Date': 'Date', 'Model': 'Model Type'},
+        markers=True
+    )
+    return fig
+
+# ==========================================
+# 4. Gradio 界面
+# ==========================================
+
+DEFAULT_GPU_FILE = "./gpu_script/gpu_price_history.csv" 
+DEFAULT_LLM_FILE = "./llm_token_tracker/llm_price_trends.csv"
+
+def load_gpu_pipeline():
+    df = clean_and_read_file(DEFAULT_GPU_FILE)
+    df = process_gpu_data(df)
+    return df, plot_gpu_trends(df)
+
+def load_llm_pipeline():
+    df = clean_and_read_file(DEFAULT_LLM_FILE)
+    df = process_llm_data(df)
+    return df, plot_llm_trends(df)
+
+# --- UI Definition ---
+with gr.Blocks(title="AI Price Tracker") as demo:
+    gr.Markdown("## 📊 AI Compute & Model Price Trends")
+    
+    with gr.Tabs():
+        # GPU Tab
+        with gr.TabItem("GPU Prices"):
+            with gr.Row():
+                with gr.Column(scale=1):
+                    gpu_plot = gr.Plot(label="Price Trend")
+            with gr.Row():
+                with gr.Accordion("Data Preview", open=False):
+                    gpu_table = gr.DataFrame()
+
+        # LLM Tab (Updated: No Filter)
+        with gr.TabItem("LLM Prices"):
+            with gr.Row():
+                # 直接展示图表，不分栏
+                with gr.Column(scale=1):
+                    llm_plot = gr.Plot(label="Price Trend")
+            
+            with gr.Row():
+                with gr.Accordion("Data Preview", open=False):
+                    llm_table = gr.DataFrame()
+
+    # --- Initialization Logic ---
+    def init_on_load():
+        # Load GPU
+        g_df, g_fig = load_gpu_pipeline()
+        
+        # Load LLM (No checkbox needed anymore)
+        l_df, l_fig = load_llm_pipeline()
+        
+        return (
+            g_fig,      # gpu_plot
+            g_df,       # gpu_table
+            l_fig,      # llm_plot
+            l_df        # llm_table
+        )
+
+    # 绑定加载事件
+    demo.load(
+        init_on_load, 
+        inputs=None, 
+        outputs=[
+            gpu_plot, 
+            gpu_table, 
+            llm_plot, 
+            llm_table
+        ]
+    )
+
+if __name__ == "__main__":
+    demo.launch(share=True)