Settings: condenser first in radio + help; flatten Model wrapper so its label aligns with the rest

app.py

@@ -667,7 +667,14 @@ gradio-app > div {
 #quest-temperature .gr-form,
 #quest-temperature .container,
 #quest-temperature .wrap-inner,
-#quest-temperature .head {
+#quest-temperature .head,
+#quest-model,
+#quest-model > div,
+#quest-model .form,
+#quest-model .gr-form,
+#quest-model .container,
+#quest-model .wrap-inner,
+#quest-model .head {
   background: transparent !important;
   border: 0 !important;
   outline: 0 !important;
@@ -1455,7 +1462,7 @@ def _trace_to_json(state: "AgentState", used_model: str) -> str:
     )
 
 
-MEMORY_STRATEGIES = ("vanilla", "condenser", "discard_all", "hide_tool_result")
+MEMORY_STRATEGIES = ("condenser", "vanilla", "discard_all", "hide_tool_result")
 
 
 def _normalize_memory_strategy(strategy: str) -> str:
@@ -1899,6 +1906,7 @@ with gr.Blocks(
                     label="Model",
                     value=QUEST_MODEL_ID,
                     interactive=False,
+                    elem_id="quest-model",
                 )
                 memory_strategy = gr.Radio(
                     label="Memory Strategy",
@@ -1908,8 +1916,8 @@ with gr.Blocks(
                 )
                 gr.HTML(
                     '<div class="memory-help">'
-                    '<b>vanilla</b> — memory management disabled; the full conversation history is kept.<br>'
                     '<b>condenser</b> (default) — when context grows large, a State Summarizer LLM compresses earlier turns into a structured JSON of trusted/untrusted/uncertain claims, visited sources, and prior search queries; the agent continues with that compact state.<br>'
+                    '<b>vanilla</b> — memory management disabled; the full conversation history is kept.<br>'
                     '<b>discard_all</b> — when context grows large, the entire message history is reset, restarting the agent from the original question with no accumulated context.<br>'
                     '<b>hide_tool_result</b> — when context grows large, older tool responses are pruned; only the most recent tool result is kept.'
                     '</div>'