Hello, I am using the VLM with Jetson Platform Services version 2.
I added an rtsp stream using the vst service and the curl commands provided in the documentation page. However, when I tried to send a delete request to delete that stream, I got the “Serve timed out processing the request” error, and the stream wasn’t getting deleted until I restarted the docker. Also, the output video became unavailable after a few minutes, and the rtsp://0.0.0.0:5011/out link didn’t stream any outputs. In the main_config file, I have changed the multi_frame_input to 8 and multi_frame_input_time to 1000. The alert_cooldown is set to 45. And the rest of the config file are unchanged.
I tried several times and I the same thing happened.
Any ideas what is the problem?
I can delete the stream in my side with JPS 2.0.1 in: Jetson Platform Services Reference Workflow & Resources | NVIDIA NGC. The doc: Visual Language Models (VLM) with Jetson Platform Services — Jetson Platform Services documentation
$ curl --location --request DELETE 'http://0.0.0.0:5010/api/v1/live-stream/b0ff6ec1-ca43-4c49-9030-1b60fa4d16dd'
{"detail":"Stream removed successfully"}
Thanks for the reply, I checked the log and apparently the video streaming face some issue. I get “FMO streams not supported for T210 onwards” log and then the Opening in BLOCKING MODE. And this blocks the output and cause an internal server error, which doesn’t let the Delete request go through.
Also, I noticed that the in the recent changes in the Jetson Platform services Docs, it is stated that the Alert functionality is using a Websocket (“True and False states as determined by the VLM, are sent to a websocket and the jetson monitoring service”), however in the implementations it is still using the REST APIs endpoints. Are the REST API endpoints for the Alerts going to the deprecated at some point?
Can you test with VST supported cameras: VST — Jetson Platform Services documentation?
Which REST API are you using? You can check all supported API in here: Visual Language Models (VLM) with Jetson Platform Services — Jetson Platform Services documentation
I am using the same REST APIs provided in the VLM with Jetson Platform Services documentation.
Visual Language Models (VLM) with Jetson Platform Services — Jetson Platform Services documentation
However, in the documentation, it is said that the alerts are sent using a webSocket to Jetson Platform Service.
Web Socket Alerts (ws://0.0.0.0:5016/api/v1/alerts/ws): Web Socket that will output Alerts when the VLM determines it is True.
This contrast with the Alert implementation in the same documentation, which is using REST APIs for setting the alert.
So my question is that are the REST APIs endpoint for setting alerts going to be deprecated and replaced with webSocket? Or will they continue to work even though there is websocket for alerts?
You can use the REST APIs for setting one alert. Then you can keep monitoring the alert using websocket. The application will received message from websocket if the VLM decide is True.