Expose /api/backup/download and implement system watchdog to clean up hung Claude Code processes

backend.py

@@ -1200,6 +1200,31 @@ async def get_models_status():
     return status_list
 
 
+import shutil
+import threading
+import signal
+from fastapi.responses import FileResponse
+
+@app.get("/api/backup/download")
+async def download_backup(authorization: str = Header(None)):
+    auth(authorization)
+    archive_base = "/tmp/workspace_backup_download"
+    archive_zip = archive_base + ".zip"
+    if os.path.exists(archive_zip):
+        try:
+            os.unlink(archive_zip)
+        except Exception:
+            pass
+            
+    try:
+        shutil.make_archive(archive_base, 'zip', WORKSPACE_DIR)
+        if not os.path.exists(archive_zip):
+            raise HTTPException(status_code=500, detail="Failed to create zip archive")
+        return FileResponse(archive_zip, media_type="application/zip", filename="workspace_backup.zip")
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=str(e))
+
+
 @app.get("/health")
 async def health():
     return {
@@ -1213,6 +1238,52 @@ async def health():
     }
 
 
+# ---------------------------------------------------------------------------
+# Watchdog Daemon for Claude Code Subprocesses
+# ---------------------------------------------------------------------------
+
+def run_watchdog():
+    log_activity("System Watchdog Daemon started")
+    while True:
+        try:
+            import psutil
+            for proc in psutil.process_iter(['pid', 'name', 'cmdline', 'create_time']):
+                try:
+                    cmd = " ".join(proc.info['cmdline'] or [])
+                    if "claude-code" in cmd.lower() or "anthropic" in cmd.lower():
+                        elapsed = time.time() - proc.info['create_time']
+                        if elapsed > 600: # 10 minutes limit
+                            log_activity(f"[Watchdog SIGKILL] Reaping hung Claude Code process PID {proc.info['pid']} (Active for {elapsed:.1f}s)")
+                            proc.kill()
+                except Exception:
+                    continue
+        except ImportError:
+            # Fallback zero-dependency shell parser
+            try:
+                out = subprocess.check_output("ps -o pid,etime,args | grep -E 'claude-code|anthropic' | grep -v grep", shell=True, text=True)
+                for line in out.strip().split("\n"):
+                    parts = line.strip().split(None, 2)
+                    if len(parts) >= 2:
+                        pid = int(parts[0])
+                        etime = parts[1]
+                        # Check if running > 10 mins (format dd-hh:mm:ss or mm:ss)
+                        is_stale = "-" in etime or len(etime.split(":")) > 2 or (len(etime.split(":")) == 2 and int(etime.split(":")[0]) > 10)
+                        if is_stale:
+                            log_activity(f"[Watchdog SIGKILL Fallback] Reaping hung process PID {pid} (etime: {etime})")
+                            os.kill(pid, signal.SIGKILL)
+            except Exception:
+                pass
+        except Exception as e:
+            log_activity(f"[Watchdog Error] {e}")
+        time.sleep(60)
+
+
+@app.on_event("startup")
+async def startup_event():
+    # Start the watchdog thread on startup
+    threading.Thread(target=run_watchdog, daemon=True).start()
+
+
 # ---------------------------------------------------------------------------
 # Entrypoint
 # ---------------------------------------------------------------------------