Update app.py

app.py

@@ -1,42 +1,62 @@
 import gradio as gr
-from transformers import pipeline
+from transformers import pipeline, AutoModelForSequenceClassification, AutoTokenizer
 
-# Load a crypto-specific sentiment model (e.g., ElKulako/cryptobert)
-sentiment_pipeline = pipeline(
-    "text-classification",
-    model="ElKulako/cryptobert",  # Pre-trained on crypto data
-    tokenizer="ElKulako/cryptobert"
-)
+# Model name
+model_name = "ElKulako/cryptobert"
+
+# Load model and tokenizer explicitly to prevent errors
+try:
+    model = AutoModelForSequenceClassification.from_pretrained(model_name)
+    tokenizer = AutoTokenizer.from_pretrained(model_name)
 
+    # Load the pipeline
+    sentiment_pipeline = pipeline(
+        "text-classification",
+        model=model,
+        tokenizer=tokenizer
+    )
+except Exception as e:
+    print(f"Error loading model: {e}")
+    exit(1)
+
+# Sentiment analysis function with keyword overrides
 def analyze(text):
-    # Get the model's initial prediction
-    result = sentiment_pipeline(text)[0]
-    
-    # Override logic for crypto-specific keywords (case-insensitive)
-    text_lower = text.lower()
-    
-    # Force "positive" for bullish terms
-    bullish_keywords = ["etf approved", "bullish", "halving", "burn", "greenlighted"]
-    if any(keyword in text_lower for keyword in bullish_keywords):
-        return {"label": "positive", "score": 0.99}
+    if not text or text.strip() == "":
+        return {"error": "No text provided."}
+
+    try:
+        # Get the model's initial prediction
+        result = sentiment_pipeline(text)[0]
         
-    # Force "negative" for bearish terms
-    bearish_keywords = ["sec lawsuit", "hack", "fud", "sell-off", "delist"]
-    if any(keyword in text_lower for keyword in bearish_keywords):
-        return {"label": "negative", "score": 0.99}
-    
-    # Return original prediction if no keywords matched
-    return {"label": result["label"], "score": result["score"]}
+        # Override logic for crypto-specific keywords (case-insensitive)
+        text_lower = text.lower()
+        
+        # Force "positive" for bullish terms
+        bullish_keywords = ["etf approved", "bullish", "halving", "burn", "greenlighted"]
+        if any(keyword in text_lower for keyword in bullish_keywords):
+            return {"label": "positive", "score": 0.99}
+            
+        # Force "negative" for bearish terms
+        bearish_keywords = ["sec lawsuit", "hack", "fud", "sell-off", "delist"]
+        if any(keyword in text_lower for keyword in bearish_keywords):
+            return {"label": "negative", "score": 0.99}
+        
+        # Return original prediction if no keywords matched
+        return {"label": result["label"], "score": result["score"]}
+
+    except Exception as e:
+        return {"error": str(e)}
 
 # Configure Gradio interface for API compatibility
 gr.Interface(
     fn=analyze,
     inputs=gr.Textbox(placeholder="Enter crypto news headline..."),
-    outputs=gr.JSON(),  # JSON output for n8n integration
+    outputs=gr.JSON(),
     title="Crypto-Specific Sentiment Analysis",
     description="Analyzes sentiment of crypto news headlines. Overrides neutral predictions for key terms like 'ETF approved' or 'SEC lawsuit'.",
-    flagging_mode="never"
+    allow_flagging="never"
 ).launch(
     server_name="0.0.0.0",
-    server_port=7860
-)
+    server_port=7860,
+    enable_queue=True  # Ensures API requests are handled properly
+)