Create app.py

app.py

@@ -0,0 +1,56 @@
+import gradio as gr
+from peft import AutoPeftModelForCausalLM
+from transformers import AutoTokenizer
+from huggingface_hub import login
+import torch
+
+# Login to HF (use your READ token)
+login("YOUR_HF_READ_TOKEN_HERE")  # Replace with your token
+
+# Model setup (loads once on Space startup)
+model_id = "agarkovv/CryptoTrader-LM"
+base_model_id = "mistralai/Ministral-8B-Instruct-2410"
+MAX_LENGTH = 32768
+DEVICE = "cuda" if torch.cuda.is_available() else "cpu"  # Use GPU if available (ZeroGPU on HF)
+
+model = AutoPeftModelForCausalLM.from_pretrained(model_id)
+tokenizer = AutoTokenizer.from_pretrained(base_model_id)
+model = model.to(DEVICE)
+model.eval()
+
+def predict_trading_decision(prompt: str) -> str:
+    """Predict daily trading decision (buy, sell, or hold) for BTC or ETH based on news and historical prices.
+    
+    Args:
+        prompt: Input prompt containing cryptocurrency news and historical price data (format: [INST]YOUR PROMPT HERE[/INST]).
+    
+    Returns:
+        Generated trading decision as text (e.g., 'Buy BTC at $62k').
+    """
+    # Format prompt as required
+    formatted_prompt = f"[INST]{prompt}[/INST]"
+    
+    inputs = tokenizer(
+        formatted_prompt, return_tensors="pt", padding=False, max_length=MAX_LENGTH, truncation=True
+    )
+    inputs = {key: value.to(model.device) for key, value in inputs.items()}
+
+    res = model.generate(
+        **inputs,
+        use_cache=True,
+        max_new_tokens=MAX_LENGTH,
+    )
+    output = tokenizer.decode(res[0], skip_special_tokens=True)
+    return output
+
+# Gradio Interface
+demo = gr.Interface(
+    fn=predict_trading_decision,
+    inputs=gr.Textbox(label="Input Prompt (News + Prices)"),
+    outputs=gr.Textbox(label="Trading Decision"),
+    title="CryptoTrader-LM MCP Tool",
+    description="Predict buy/sell/hold for BTC/ETH."
+)
+
+# Launch with MCP support
+demo.launch(mcp_server=True)