feat: add multi-agent hierarchy display

agents/orchestrator.py

@@ -633,9 +633,24 @@ Return JSON:
         )
 
     def _create_step_callback(self, agent_key: str):
-        """Create a step callback for tracking agent execution."""
+        """Create a step callback for tracking agent execution (multi-agent style)."""
+        agent_started = {"value": False}
+
         def step_callback(step: ActionStep, agent) -> None:
             """Callback triggered at each agent step."""
+            # Emit agent_start on first step (shows agent as parent of its tools)
+            if not agent_started["value"]:
+                agent_started["value"] = True
+                config = self.AGENT_CONFIG.get(agent_key, {})
+                self._event_queue.put(("agent_start", {
+                    "agent": agent_key,
+                    "name": config.get("name", agent_key),
+                    "icon": config.get("icon", "🤖"),
+                    "description": config.get("description", ""),
+                    "start_time": time.time(),
+                }))
+                print(f"Starting {config.get('name', agent_key)}...")
+
             step_data = {
                 "agent": agent_key,
                 "step_number": getattr(step, 'step_number', 0),

app.py

@@ -188,38 +188,52 @@ def _extract_tool_input_preview(tool_name: str, input_str: str) -> str:
 def build_inline_progress(agent_states: dict, elapsed: float, current_activity: str = "",
                           thinking: str = "", tool_history: list = None) -> str:
     """
-    Build Claude Code-style progress display.
+    Build multi-agent style progress display.
 
     Visual pattern:
-    ● Tool call (action)
-      ⎿ Result (indented)
-    ∴ Thinking... (analysis)
+    ● Agent Name
+      ├── Tool call...
+      │   ⎿  Result
+      ├── Tool call...
+      │   ⎿  Result
     """
     lines = []
 
     # Planning header (shown at start)
     if elapsed < 3 and not tool_history:
-        lines.append("🤖 **Planning:** Research → Report\n")
+        lines.append("🤖 **Controller planning...**\n")
 
-    # Tool calls with results (Claude Code style)
+    # Group tools by agent (multi-agent style)
     if tool_history:
+        current_agent = None
+        agent_icons = {"triage": "🎯", "research": "🔍", "report": "📋"}
+        agent_names = {"triage": "Triage Agent", "research": "Research Agent", "report": "Report Agent"}
+
         for tool in tool_history:
+            agent = tool.get("agent")
             name = tool.get("display_name", tool.get("tool", "Unknown"))
             status = tool.get("status", "running")
             result = tool.get("result_summary", "")
             input_preview = tool.get("input_preview", "")
 
-            # Tool call line
+            # Show agent header when agent changes
+            if agent and agent != current_agent:
+                current_agent = agent
+                icon = agent_icons.get(agent, "🤖")
+                agent_name = agent_names.get(agent, agent.title())
+                lines.append(f"● {icon} **{agent_name}**")
+
+            # Tool call line (indented under agent)
             if input_preview:
-                lines.append(f"● {name}({input_preview})")
+                lines.append(f"  ├── {name}({input_preview})")
             else:
-                lines.append(f"● {name}...")
+                lines.append(f"  ├── {name}...")
 
-            # Result line (indented)
+            # Result line (further indented)
             if status == "done" and result:
-                lines.append(f"  ⎿  {result}")
+                lines.append(f"  │   ⎿  {result}")
             elif status == "done":
-                lines.append(f"  ⎿  Done")
+                lines.append(f"  │   ⎿  Done")
 
     # Thinking/analysis section
     if thinking: