fix(crew): prevent zombie workers and improve TUI robustness

ghostcrew/agents/crew/models.py

@@ -20,6 +20,7 @@ class AgentStatus(Enum):
     PENDING = "pending"
     RUNNING = "running"
     COMPLETE = "complete"
+    WARNING = "warning"  # Completed but hit max iterations
     ERROR = "error"
     CANCELLED = "cancelled"
 

ghostcrew/agents/crew/worker_pool.py

@@ -109,16 +109,20 @@ class WorkerPool:
         worker_runtime = LocalRuntime()
         await worker_runtime.start()
 
+        from ...config.constants import WORKER_MAX_ITERATIONS
+
         agent = GhostCrewAgent(
             llm=self.llm,
             tools=self.tools,
             runtime=worker_runtime,  # Use isolated runtime
             target=self.target,
             rag_engine=self.rag_engine,
+            max_iterations=WORKER_MAX_ITERATIONS,
         )
 
         try:
             final_response = ""
+            hit_max_iterations = False
             async for response in agent.agent_loop(worker.task):
                 # Track tool calls
                 if response.tool_calls:
@@ -136,19 +140,34 @@ class WorkerPool:
                 # Capture final response (text without tool calls)
                 if response.content and not response.tool_calls:
                     final_response = response.content
+                
+                # Check if max iterations was hit
+                if response.metadata and response.metadata.get("max_iterations_reached"):
+                    hit_max_iterations = True
 
             worker.result = final_response or "No findings."
-            worker.status = AgentStatus.COMPLETE
             worker.completed_at = time.time()
             self._results[worker.id] = worker.result
 
-            self._emit(
-                worker.id,
-                "complete",
-                {
-                    "summary": worker.result[:200],
-                },
-            )
+            if hit_max_iterations:
+                worker.status = AgentStatus.WARNING
+                self._emit(
+                    worker.id,
+                    "warning",
+                    {
+                        "summary": worker.result[:200],
+                        "reason": "Max iterations reached",
+                    },
+                )
+            else:
+                worker.status = AgentStatus.COMPLETE
+                self._emit(
+                    worker.id,
+                    "complete",
+                    {
+                        "summary": worker.result[:200],
+                    },
+                )
 
         except asyncio.CancelledError:
             worker.status = AgentStatus.CANCELLED

ghostcrew/config/constants.py

@@ -57,6 +57,7 @@ DEFAULT_MAX_TOKENS = 4096
 
 # Agent Defaults
 DEFAULT_MAX_ITERATIONS = int(os.environ.get("GHOSTCREW_MAX_ITERATIONS", "50"))
+WORKER_MAX_ITERATIONS = int(os.environ.get("GHOSTCREW_WORKER_MAX_ITERATIONS", "10"))
 
 # File Extensions
 KNOWLEDGE_TEXT_EXTENSIONS = [".txt", ".md"]

ghostcrew/interface/tui.py

@@ -1310,7 +1310,7 @@ Be concise. Use the actual data from notes."""
         wtype = worker.get("worker_type", "worker")
         findings = worker.get("findings", 0)
 
-        # Simple 3-state icons: working (braille), done (checkmark), error (X)
+        # 4-state icons: working (braille), done (checkmark), warning (!), error (X)
         if status in ("running", "pending"):
             # Animated braille spinner for all in-progress states
             icon = self._spinner_frames[self._spinner_frame % len(self._spinner_frames)]
@@ -1318,6 +1318,9 @@ Be concise. Use the actual data from notes."""
         elif status == "complete":
             icon = "✓"
             color = "#22c55e"  # green
+        elif status == "warning":
+            icon = "!"
+            color = "#f59e0b"  # amber/orange
         else:  # error, cancelled, unknown
             icon = "✗"
             color = "#ef4444"  # red
@@ -1360,6 +1363,14 @@ Be concise. Use the actual data from notes."""
                 )
                 self._crew_findings_count += findings_count
                 self._update_crew_stats()
+            elif event_type == "warning":
+                # Worker hit max iterations but has results
+                self._update_crew_worker(worker_id, status="warning")
+                reason = data.get("reason", "Partial completion")
+                worker = self._crew_workers.get(worker_id, {})
+                wtype = worker.get("worker_type", "worker")
+                self._add_system(f"[!] {wtype.upper()} stopped: {reason}")
+                self._update_crew_stats()
             elif event_type == "error":
                 self._update_crew_worker(worker_id, status="error")
                 worker = self._crew_workers.get(worker_id, {})
@@ -1671,6 +1682,14 @@ Be concise. Use the actual data from notes."""
             self._is_running = False
 
     def action_quit_app(self) -> None:
+        # Stop any running tasks first
+        if self._is_running:
+            self._should_stop = True
+            if self._current_worker and not self._current_worker.is_finished:
+                self._current_worker.cancel()
+            if self._current_crew:
+                # Schedule cancel but don't wait - we're exiting
+                asyncio.create_task(self._cancel_crew())
         self.exit()
 
     def action_stop_agent(self) -> None:
@@ -1682,6 +1701,10 @@ Be concise. Use the actual data from notes."""
             if self._current_worker and not self._current_worker.is_finished:
                 self._current_worker.cancel()
 
+            # Cancel crew orchestrator if running
+            if self._current_crew:
+                asyncio.create_task(self._cancel_crew())
+
             # Clean up agent state to prevent stale tool responses
             if self.agent:
                 self.agent.cleanup_after_cancel()
@@ -1690,6 +1713,19 @@ Be concise. Use the actual data from notes."""
             if self.mcp_manager:
                 asyncio.create_task(self._reconnect_mcp_after_cancel())
 
+    async def _cancel_crew(self) -> None:
+        """Cancel crew orchestrator and all workers."""
+        try:
+            if self._current_crew:
+                await self._current_crew.cancel()
+                self._current_crew = None
+                # Mark all running workers as cancelled in the UI
+                for worker_id, worker in self._crew_workers.items():
+                    if worker.get("status") in ("running", "pending"):
+                        self._update_crew_worker(worker_id, status="cancelled")
+        except Exception:
+            pass  # Best effort
+
     async def _reconnect_mcp_after_cancel(self) -> None:
         """Reconnect MCP servers after cancellation to restore clean state."""
         await asyncio.sleep(0.5)  # Brief delay for cancellation to propagate