better error handling, try to avoid segfault in sillytavern

koboldcpp.py

@@ -568,6 +568,7 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
         current_token = 0
         incomplete_token_buffer = bytearray()
         await asyncio.sleep(0.05) #anti race condition, prevent check from overtaking generate
+        try:
             while True:
                 streamDone = handle.has_finished() #exit next loop on done
                 tokenStr = ""
@@ -605,6 +606,9 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
                     if api_format == 4:  # if oai chat, send last [DONE] message consistent with openai format
                         await self.send_oai_sse_event('[DONE]')
                     break
+        except Exception as ex:
+            print("SSE streaming was interrupted due to an exception")
+            print(ex)
 
         # flush buffers, sleep a bit to make sure all data sent, and then force close the connection
         self.wfile.flush()
@@ -629,7 +633,7 @@ class ServerRequestHandler(http.server.SimpleHTTPRequestHandler):
         except (BrokenPipeError, ConnectionAbortedError) as cae: # attempt to abort if connection lost
             print(cae)
             handle.abort_generate()
-            time.sleep(0.1) #short delay
+            time.sleep(0.2) #short delay
         except Exception as e:
             print(e)
 
@@ -938,8 +942,10 @@ Enter Prompt:<br>
                         self.send_header('content-length', str(len(genresp)))
                         self.end_headers(content_type='application/json')
                         self.wfile.write(genresp)
-                except:
+                except Exception as ex:
                     print("Generate: The response could not be sent, maybe connection was terminated?")
+                    handle.abort_generate()
+                    time.sleep(0.2) #short delay
                 return
         finally:
             modelbusy.release()