perf: allow disabling websocket per-message-deflate (#28613)

With delta streaming most websocket frames are tiny per-token deltas, and
per-message-deflate pays zlib work on every outgoing frame per subscriber for
near-zero gain there; under heavy streaming that shows up as measurable server
CPU. The frames that still benefit are the rare large ones (final message,
sources), and even a 100k token message is only a few hundred KB uncompressed,
which any network delivers without noticeable delay.

UVICORN_WS_PER_MESSAGE_DEFLATE=false (default true, current behavior) disables
the extension in every entry point: open-webui serve and dev, start.sh both
invocations, start_windows.bat and dev.sh. Verified against a running
instance: with the flag off the server declines the client-offered
permessage-deflate extension, with defaults it still negotiates it.

backend/open_webui/__init__.py

@@ -74,7 +74,7 @@ def serve(
             os.environ['LD_LIBRARY_PATH'] = ':'.join(LD_LIBRARY_PATH)
 
     import open_webui.main  # noqa: F401
-    from open_webui.env import UVICORN_WORKERS  # Import the workers setting
+    from open_webui.env import UVICORN_WORKERS, UVICORN_WS_PER_MESSAGE_DEFLATE
 
     # On Windows, uvicorn's default loop factory hardcodes ProactorEventLoop,
     # which is incompatible with psycopg v3 async.  Setting loop='none' lets
@@ -87,6 +87,7 @@ def serve(
         port=port,
         forwarded_allow_ips='*',
         workers=UVICORN_WORKERS,
+        ws_per_message_deflate=UVICORN_WS_PER_MESSAGE_DEFLATE,
         loop=loop,
     )
 
@@ -97,12 +98,15 @@ def dev(
     port: int = 8080,
     reload: bool = True,
 ):
+    from open_webui.env import UVICORN_WS_PER_MESSAGE_DEFLATE
+
     uvicorn.run(
         'open_webui.main:app',
         host=host,
         port=port,
         reload=reload,
         forwarded_allow_ips='*',
+        ws_per_message_deflate=UVICORN_WS_PER_MESSAGE_DEFLATE,
     )
 
 

backend/open_webui/env.py

@@ -443,6 +443,9 @@ try:
 except (ValueError, TypeError):
     UVICORN_WORKERS = 1
 
+# tiny delta-stream frames make per-frame websocket compression CPU-bound, allow opting out (true/false)
+UVICORN_WS_PER_MESSAGE_DEFLATE = os.getenv('UVICORN_WS_PER_MESSAGE_DEFLATE', 'True').lower() == 'true'
+
 ####################################
 # WEBSOCKET SUPPORT
 ####################################