Fix: Switch from gated Llama 3.2 to ungated TinyLlama-1.1B-Chat

app.py

@@ -495,23 +495,23 @@ class ExponentialLearner:
         try:
             # Step 1: Initialize model (cached after first load)
             if not hasattr(self, 'local_ai'):
-                logger.info("🔄 Step 1/5: Loading Llama 3.2-1B-Instruct model...")
+                logger.info("🔄 Step 1/5: Loading TinyLlama-1.1B-Chat model...")
                 try:
                     from transformers import pipeline
                     logger.info("   ✅ Transformers library imported")
                     
                     self.local_ai = pipeline(
                         "text-generation",
-                        model="meta-llama/Llama-3.2-1B-Instruct",
+                        model="TinyLlama/TinyLlama-1.1B-Chat-v1.0",
                         device="cpu",
                         torch_dtype="auto"
                     )
-                    logger.info("   ✅ Llama 3.2 model loaded successfully!")
+                    logger.info("   ✅ TinyLlama model loaded successfully!")
                 except Exception as model_error:
                     logger.error(f"   ❌ Model loading FAILED: {model_error}", exc_info=True)
                     raise Exception(f"AI Model Loading Failed: {model_error}")
             else:
-                logger.info("✅ Step 1/5: Using cached Llama 3.2 model")
+                logger.info("✅ Step 1/5: Using cached TinyLlama model")
             
             # Step 2: Construct structured AI prompt with clear output format
             logger.info("📝 Step 2/5: Constructing trading prompt...")