Update app.py

app.py

@@ -48,42 +48,12 @@ Here is example of your workflow. This example consists of your multiple respons
 
 **Example Workflow: User Task - "Tell me the latest AI news"**  
 
-### **Phase 1: Initial Search**  
-- **Action**: Clarification of the request is not necessary. Agent formulates **broad search queries** to capture high-level developments.  
-  - Example queries:  
-    ```  
-    <search>  
-    "latest AI news 2023"  
-    "recent AI breakthroughs"  
-    "AI advancements October 2023"  
-    "top AI research papers this month"  
-    </search>  
-    ```  
-- **Then wait for web search results**: Simulated outputs from credible sources (e.g., TechCrunch, Reuters, MIT Tech Review, ArXiv).  
+1. **Search Process**: `Initial broad queries → Identify key themes → Sequential targeted sub-searches (one <search> block per theme and one response per <search>) → Verify all data through iterative steps.`  
+2. **Answer Synthesis**: `Structure findings into detailed sections (technical specs, comparisons, limitations. Ensure 300+ words with subheadings.) → Cite all sources.`
 
 ---
 
-### **Phase 2 and more: Follow-Up Searches**  
-- **Action**: Agent identifies **key themes** from initial results (e.g., AlphaCode 2, EU regulations, healthcare AI, Llama 3) and conducts **targeted sub-searches** for each.  
-  - **Sub-Search Structure**:  
-    1. **AlphaCode 2**: Technical specs, training data, comparisons, limitations.  
-    2. **EU AI Ethics Guidelines**: Regulatory specifics, public reaction, comparisons to other regions.  
-    3. **Healthcare AI**: Dataset details, peer reviews, clinical trial plans.  
-    4. **Llama 3**: Benchmarks, applications, limitations vs. competitors.  
-  - **Note**: Each sub-search uses a new `<search>` block and new response with **specific queries** (e.g., "AlphaCode 2 training data sources").  
-
----
-
-### **Final Phase: Final Answer**  
-- **Action**: Agent synthesizes findings into a **structured report** with:  
-  - **Exhaustive technical details** (e.g., dataset sizes, validation methods).  
-  - **Multi-step analysis** (e.g., comparing AI performance metrics).  
-  - **Subheadings** (e.g., "Technical Breakthroughs," "Regulatory Impacts").  
-- **Output**: A **300+ word answer** citing all sources, formatted with bullet points and clear attribution.  
-
----
-
-This was an example of your workflow, this is not your single response. You can use <search> command only once per response.
+This was an example of your workflow, this is NOT your single response. You can use <search> command only once per response.
 
 **Termination Conditions:**
 - Exhaust all logical search avenues before finalizing answers.