Update app.py

app.py

@@ -50,16 +50,47 @@ def analyze_world_model(model_name, dataset_key, num_samples=25):
     state_assignments = kmeans.labels_
     
     # Step C: State Elaboration Logic
-    state_info = "### 🧠 State Interpretation & Dataset Mapping\n"
     cluster_texts = collections.defaultdict(list)
     for idx, cluster_id in enumerate(state_assignments):
         cluster_texts[cluster_id].append(input_snippets[idx])
-        
+
+    # Initialize Gemini
+    gemini_model = genai.GenerativeModel('gemini-1.5-flash')
+    state_info = "## 🧠 Newtonian State Interpretation\n"
+
     for cluster_id in range(n_clusters):
         snippets = cluster_texts[cluster_id]
-        # Identify common tokens/attributes that represent this state
-        summary = " | ".join([s[:40] + "..." for s in snippets[:2]])
-        state_info += f"**State S{cluster_id}**: Representing context such as: *{summary}*\n\n"
+        context_payload = "\n".join([f"- {s}" for s in snippets[:4]])
+        
+        # Proper prompt engineering to decode the 'Equivalence Class'
+        prompt = f"""
+        Analyze these text snippets from the '{dataset_key}' dataset that fall into the same latent state cluster.
+        Identify the CORE structural or semantic theme (e.g., 'Historical Narrative', 'Technical Development', 'Numerical Lists').
+        
+        Text Snippets:
+        {context_payload}
+        
+        Format your response exactly as:
+        **State S{cluster_id} [Label]**: [One sentence explanation of the shared logic/context].
+        """
+        
+        try:
+            response = gemini_model.generate_content(prompt)
+            state_info += response.text.strip() + "\n\n"
+        except Exception:
+            # Fallback if API fails
+            summary = " | ".join([s[:40] + "..." for s in snippets[:2]])
+            state_info += f"**State S{cluster_id}**: Context: *{summary}*\n\n"
+    # state_info = "### 🧠 State Interpretation & Dataset Mapping\n"
+    # cluster_texts = collections.defaultdict(list)
+    # for idx, cluster_id in enumerate(state_assignments):
+    #     cluster_texts[cluster_id].append(input_snippets[idx])
+        
+    # for cluster_id in range(n_clusters):
+    #     snippets = cluster_texts[cluster_id]
+    #     # Identify common tokens/attributes that represent this state
+    #     summary = " | ".join([s[:40] + "..." for s in snippets[:2]])
+    #     state_info += f"**State S{cluster_id}**: Representing context such as: *{summary}*\n\n"
 
     # Step D: DFA Reconstruction
     G = nx.DiGraph()