Prevent Quart from cancelling active chat streams after 60 seconds
This commit is contained in:
@@ -619,7 +619,7 @@ async def stream_query():
|
||||
_log_summary()
|
||||
yield "data: [DONE]\n\n"
|
||||
|
||||
return await make_response(
|
||||
response = await make_response(
|
||||
event_generator(),
|
||||
200,
|
||||
{
|
||||
@@ -628,6 +628,10 @@ async def stream_query():
|
||||
"X-Accel-Buffering": "no",
|
||||
},
|
||||
)
|
||||
# Quart's default 60s response deadline cancels even active SSE streams.
|
||||
# Silent model calls are bounded separately by MODEL_STALL_TIMEOUT.
|
||||
response.timeout = None
|
||||
return response
|
||||
|
||||
|
||||
@conversation_blueprint.route("/<conversation_id>")
|
||||
|
||||
Reference in New Issue
Block a user