Update app.py

app.py

@@ -302,17 +302,21 @@ You have access to the following tools:
 * Do not add any text before or after the JSON block.
 
 **REASONING PROCESS & STOPPING CONDITION:**
-1.  Analyze the question.
-2.  Break down problem into steps.
-3.  Determine if tools are needed.
-4.  If needed, call tool using JSON format. Wait for output.
-5.  Analyze tool output.
+1.  **Analyze:** Read the user's question carefully. Identify the goal.
+2.  **Plan:** Break down the problem into a sequence of logical steps. 
+    * **For logic puzzles (like riddles):** You *must* use the `code_interpreter`. Your *first* step is to write out your step-by-step logical reasoning as Python comments *before* writing any solving code.
+3.  **Execute:** Choose the *one* best tool for the *next* step in your plan.
+4.  **Call Tool:** Call the chosen tool using the exact JSON format. Do not add any other text.
+5.  **Analyze Output:** Receive the tool's output (either a success message or an error).
 6.  **DECISION POINT:**
-    a. **If the tool output directly contains the final answer (or allows trivial calculation):** Your *next* response MUST be a call to `final_answer_tool` with only that answer. **DO NOT** call any more tools.
-    b. **If more steps or tools are needed:** Continue reasoning, go back to step 4.
-    c. **If a tool call results in an error:** **Do not** try the exact same tool call again. Re-evaluate, try different arguments, or a different tool. If stuck, call `final_answer_tool` with an error message.
-7.  **FINAL OUTPUT:** The graph will stop *only* when you call `final_answer_tool`.
-"""
+    a. **Answer Found?** If the tool output directly contains the final answer (or makes it trivial to calculate), your *very next* action MUST be to call `final_answer_tool` with that answer.
+    b. **Error Occurred?** If the tool returned an error message:
+        * **STOP.** Do not try the *exact same tool call* again.
+        * **Rethink:** Analyze the error message. What went wrong?
+        * **Correct:** Try a *different* tool, *different* arguments, or a *different approach* entirely.
+        * If you are completely stuck after an error, call `final_answer_tool` with a concise error message (e.g., "Error: Could not find user").
+    c. **More Steps Needed?** If the answer is not yet found and there was no error, go back to step 2 (Plan) to determine the next logical step.
+7.  **FINAL OUTPUT:** The graph will stop *only* when you call `final_answer_tool`. Do not provide the answer in any other way."""
         # =============================================================
 
         print("Initializing Groq LLM Endpoint...")