Upload llm.py

src/apps/utils/llm.py

@@ -93,8 +93,9 @@ Chat History: {chat_history}
             "Set OPENROUTER_API_KEY in src/apps/.env (get a key at https://openrouter.ai/keys)"
         )
 
-    # Switched to Llama 3.2 3B as primary to avoid 429 Rate Limits
-    primary_model = "meta-llama/llama-3.2-3b-instruct:free"
+    # Switched to Mistral Small 3.1 to avoid "PROHIBITED_CONTENT" errors
+    # Mistral is much better for professional/legal reasoning without over-censorship.
+    primary_model = "mistralai/mistral-small-3.1-24b-instruct:free"
     
     # Initialize client with recommended OpenRouter headers
     client = openai.OpenAI(
@@ -106,12 +107,11 @@ Chat History: {chat_history}
         }
     )
     
-    # Reordered fallbacks to put stable models first
+    # Reordered fallbacks
     fallback_models = [
-        "google/gemma-3-4b-it:free",
         "qwen/qwen3-coder:free",
-        "google/gemma-3-12b-it:free",
-        "google/gemma-3-27b-it:free"
+        "meta-llama/llama-3.2-3b-instruct:free",
+        "google/gemma-3-4b-it:free"
     ]
 
     current_model = primary_model