Latency logging

backend/danswer/configs/app_configs.py

@@ -251,6 +251,7 @@ LOG_ALL_MODEL_INTERACTIONS = (
 LOG_VESPA_TIMING_INFORMATION = (
     os.environ.get("LOG_VESPA_TIMING_INFORMATION", "").lower() == "true"
 )
+LOG_ENDPOINT_LATENCY = os.environ.get("LOG_ENDPOINT_LATENCY", "").lower() == "true"
 # Anonymous usage telemetry
 DISABLE_TELEMETRY = os.environ.get("DISABLE_TELEMETRY", "").lower() == "true"
 

backend/danswer/main.py

@@ -27,6 +27,7 @@ from danswer.configs.app_configs import APP_PORT
 from danswer.configs.app_configs import AUTH_TYPE
 from danswer.configs.app_configs import DISABLE_GENERATIVE_AI
 from danswer.configs.app_configs import DISABLE_INDEX_UPDATE_ON_SWAP
+from danswer.configs.app_configs import LOG_ENDPOINT_LATENCY
 from danswer.configs.app_configs import OAUTH_CLIENT_ID
 from danswer.configs.app_configs import OAUTH_CLIENT_SECRET
 from danswer.configs.app_configs import USER_AUTH_SECRET
@@ -70,6 +71,7 @@ from danswer.server.manage.llm.api import basic_router as llm_router
 from danswer.server.manage.secondary_index import router as secondary_index_router
 from danswer.server.manage.slack_bot import router as slack_bot_management_router
 from danswer.server.manage.users import router as user_router
+from danswer.server.middleware.latency_logging import add_latency_logging_middleware
 from danswer.server.query_and_chat.chat_backend import router as chat_router
 from danswer.server.query_and_chat.query_backend import (
     admin_router as admin_query_router,
@@ -352,6 +354,8 @@ def get_application() -> FastAPI:
         allow_methods=["*"],
         allow_headers=["*"],
     )
+    if LOG_ENDPOINT_LATENCY:
+        add_latency_logging_middleware(application, logger)
 
     # Ensure all routes have auth enabled or are explicitly marked as public
     check_router_auth(application)

backend/danswer/server/middleware/latency_logging.py

@@ -0,0 +1,23 @@
+import logging
+import time
+from collections.abc import Awaitable
+from collections.abc import Callable
+
+from fastapi import FastAPI
+from fastapi import Request
+from fastapi import Response
+
+
+def add_latency_logging_middleware(app: FastAPI, logger: logging.LoggerAdapter) -> None:
+    @app.middleware("http")
+    async def log_latency(
+        request: Request, call_next: Callable[[Request], Awaitable[Response]]
+    ) -> Response:
+        start_time = time.monotonic()
+        response = await call_next(request)
+        process_time = time.monotonic() - start_time
+        logger.info(
+            f"Path: {request.url.path} - Method: {request.method} - "
+            f"Status Code: {response.status_code} - Time: {process_time:.4f} secs"
+        )
+        return response

deployment/data/nginx/app.conf.template

@@ -1,3 +1,9 @@
+# Log format to include request latency
+log_format custom_main '$remote_addr - $remote_user [$time_local] "$request" '
+                '$status $body_bytes_sent "$http_referer" '
+                '"$http_user_agent" "$http_x_forwarded_for" '
+                'rt=$request_time';
+
 upstream api_server {
     # fail_timeout=0 means we always retry an upstream even if it failed
     # to return a good HTTP response
@@ -20,6 +26,8 @@ server {
 
     client_max_body_size 5G;    # Maximum upload size
 
+    access_log /var/log/nginx/access.log custom_main;
+
     # Match both /api/* and /openapi.json in a single rule
     location ~ ^/(api|openapi.json)(/.*)?$ {
         # Rewrite /api prefixed matched paths

deployment/data/nginx/app.conf.template.dev

@@ -1,3 +1,9 @@
+# Override log format to include request latency
+log_format custom_main '$remote_addr - $remote_user [$time_local] "$request" '
+                      '$status $body_bytes_sent "$http_referer" '
+                      '"$http_user_agent" "$http_x_forwarded_for" '
+                      'rt=$request_time';
+
 upstream api_server {
     # fail_timeout=0 means we always retry an upstream even if it failed
     # to return a good HTTP response
@@ -20,6 +26,8 @@ server {
 
     client_max_body_size 5G;    # Maximum upload size    
 
+    access_log /var/log/nginx/access.log custom_main;
+
     # Match both /api/* and /openapi.json in a single rule
     location ~ ^/(api|openapi.json)(/.*)?$ {
         # Rewrite /api prefixed matched paths
@@ -58,3 +66,4 @@ server {
         proxy_pass http://web_server;
     }
 }
+

deployment/data/nginx/app.conf.template.no-letsencrypt

@@ -1,3 +1,9 @@
+# Log format to include request latency
+log_format custom_main '$remote_addr - $remote_user [$time_local] "$request" '
+                '$status $body_bytes_sent "$http_referer" '
+                '"$http_user_agent" "$http_x_forwarded_for" '
+                'rt=$request_time';
+
 upstream api_server {
     # fail_timeout=0 means we always retry an upstream even if it failed
     # to return a good HTTP response
@@ -20,6 +26,8 @@ server {
 
     client_max_body_size 5G;    # Maximum upload size
 
+    access_log /var/log/nginx/access.log custom_main;
+
     # Match both /api/* and /openapi.json in a single rule
     location ~ ^/(api|openapi.json)(/.*)?$ {
         # Rewrite /api prefixed matched paths

web/src/app/chat/ChatPage.tsx

@@ -337,7 +337,8 @@ export function ChatPage({
           )
         : undefined
   );
-  const livePersona = selectedPersona || filteredAssistants[0];
+  const livePersona =
+    selectedPersona || filteredAssistants[0] || availablePersonas[0];
 
   const [chatSessionSharedStatus, setChatSessionSharedStatus] =
     useState<ChatSessionSharedStatus>(ChatSessionSharedStatus.Private);