From bf08835d6f3147f151f825297483dc01c4267aa1 Mon Sep 17 00:00:00 2001
From: Classic298 <27028174+Classic298@users.noreply.github.com>
Date: Tue, 25 Aug 2026 00:46:07 +0200
Subject: [PATCH] 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/dev.sh                 | 2 +-
 backend/open_webui/__init__.py | 6 +++++-
 backend/open_webui/env.py      | 3 +++
 backend/start.sh               | 4 ++--
 backend/start_windows.bat      | 3 ++-
 5 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/backend/dev.sh b/backend/dev.sh
index 838b93f653..9737b3d005 100755
--- a/backend/dev.sh
+++ b/backend/dev.sh
@@ -1,3 +1,3 @@
 export CORS_ALLOW_ORIGIN="http://localhost:5173;http://localhost:8080"
 PORT="${PORT:-8080}"
-uvicorn open_webui.main:app --port $PORT --host 0.0.0.0 --forwarded-allow-ips "${FORWARDED_ALLOW_IPS:-*}" --reload
+uvicorn open_webui.main:app --port $PORT --host 0.0.0.0 --forwarded-allow-ips "${FORWARDED_ALLOW_IPS:-*}" --ws-per-message-deflate "${UVICORN_WS_PER_MESSAGE_DEFLATE:-true}" --reload
diff --git a/backend/open_webui/__init__.py b/backend/open_webui/__init__.py
index 82f4e5e7ba..e803cea466 100644
--- a/backend/open_webui/__init__.py
+++ b/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,
     )
 
 
diff --git a/backend/open_webui/env.py b/backend/open_webui/env.py
index 2d610a3388..34a2df1203 100644
--- a/backend/open_webui/env.py
+++ b/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
 ####################################
diff --git a/backend/start.sh b/backend/start.sh
index 0846273b89..634f4bcc1b 100755
--- a/backend/start.sh
+++ b/backend/start.sh
@@ -72,7 +72,7 @@ if [[ -n "${SPACE_ID:-}" ]]; then
   if [[ -n "${ADMIN_USER_EMAIL:-}" && -n "${ADMIN_USER_PASSWORD:-}" ]]; then
     echo "Creating admin user for Space..."
     WEBUI_SECRET_KEY="${WEBUI_SECRET_KEY:-}" \
-      uvicorn open_webui.main:app --host "$HOST" --port "$PORT" --forwarded-allow-ips "${FORWARDED_ALLOW_IPS:-*}" &
+      uvicorn open_webui.main:app --host "$HOST" --port "$PORT" --forwarded-allow-ips "${FORWARDED_ALLOW_IPS:-*}" --ws-per-message-deflate "${UVICORN_WS_PER_MESSAGE_DEFLATE:-true}" &
     webui_pid=$!
 
     echo "Waiting for server to become healthy..."
@@ -102,7 +102,7 @@ UVICORN_WORKERS="${UVICORN_WORKERS:-1}"
 if [[ "$#" -gt 0 ]]; then
   ARGS=("$@")
 else
-  ARGS=(--workers "$UVICORN_WORKERS")
+  ARGS=(--workers "$UVICORN_WORKERS" --ws-per-message-deflate "${UVICORN_WS_PER_MESSAGE_DEFLATE:-true}")
 fi
 
 exec env WEBUI_SECRET_KEY="${WEBUI_SECRET_KEY:-}" \
diff --git a/backend/start_windows.bat b/backend/start_windows.bat
index d5c4510baa..4a57d34694 100644
--- a/backend/start_windows.bat
+++ b/backend/start_windows.bat
@@ -55,5 +55,6 @@ IF "%WEBUI_SECRET_KEY% %WEBUI_JWT_SECRET_KEY%" == " " (
 :: Execute uvicorn
 SET "WEBUI_SECRET_KEY=%WEBUI_SECRET_KEY%"
 IF "%UVICORN_WORKERS%"=="" SET UVICORN_WORKERS=1
-uvicorn open_webui.main:app --host "%HOST%" --port "%PORT%" --forwarded-allow-ips %FORWARDED_ALLOW_IPS% --workers %UVICORN_WORKERS% --ws auto
+if "%UVICORN_WS_PER_MESSAGE_DEFLATE%" == "" set "UVICORN_WS_PER_MESSAGE_DEFLATE=true"
+uvicorn open_webui.main:app --host "%HOST%" --port "%PORT%" --forwarded-allow-ips %FORWARDED_ALLOW_IPS% --workers %UVICORN_WORKERS% --ws auto --ws-per-message-deflate %UVICORN_WS_PER_MESSAGE_DEFLATE%
 :: For ssl user uvicorn open_webui.main:app --host "%HOST%" --port "%PORT%" --forwarded-allow-ips '*' --ssl-keyfile "key.pem" --ssl-certfile "cert.pem" --ws auto