Failed to load CUDA modules. maybe the NVIDIA driver is not installed?
Failed to initialize CUDA context.
Failed to open VAAPI display.
Failed to open VA display. Maybe the AMD video driver or libva-dev/libdrm-dev is not installed?
Failed to load CUDA modules. maybe the NVIDIA driver is not installed?
Failed to initialize CUDA context.

why i am geting these failed errors on stating livekit agent server

Thanks for reaching out, jitendra. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

i am running server on azure container instances

The errors you're seeing seem to point to some challenges with GPU or hardware acceleration support in your Azure Container Instances (ACI) environment. ACI does have some limitations when it comes to GPU support, and features like CUDA or VAAPI might need extra configuration to work smoothly. Let’s walk through a few steps that might help: • Verify GPU Support: It’s a good idea to check if your Azure region and subscription actually support GPU-enabled ACI. If not, exploring options like Azure Kubernetes Service (AKS) with GPU nodes or Azure Virtual Machines could be a great alternative. • Use Proper Base Images: For CUDA, make sure you’re using a CUDA-enabled Docker image or building one that includes the necessary drivers. VAAPI might not be fully supported in ACI, so you might want to consider software-based alternatives for video encoding or decoding. • Check Logs: Taking a look at the container logs can reveal if there are any missing libraries or driver errors. Running commands like

livekit-agent logs

can help pinpoint specific issues. • Alternative Deployment: If you keep running into ACI limitations, deploying on AKS or a self-hosted Kubernetes cluster with GPU support might be the way to go. Could you let me know if your deployment specifically needs GPU acceleration for video processing or AI/ML tasks? That way, I can offer more tailored advice to help you out. Sources: docs/build.md | webrtc-sys/src/vaapi/vaapi_display_win32.cpp | Agent Deployment & Docker Configuration | livekit/livekit_composite | DeepW... | Agent deployment CLI reference | LiveKit Docs

my ACI running on CPU not the GPU

The errors you're seeing are probably because the LiveKit agent server is trying to start GPU-based features like CUDA or VAAPI, which unfortunately don’t work in CPU-only environments. Here are a few steps that might help you fix this: • Disable GPU Features: Make sure the LiveKit agent is set up to run in CPU-only mode. Take a look at any environment variables or configuration settings that might be turning on GPU features (like CUDA or VAAPI) and switch them off. • Use a CPU-Compatible Image: If you're working with a custom Docker image, double-check that it’s designed for CPU-only setups without any GPU dependencies. • Inspect Logs: Have a look at the container logs to spot any specific modules or libraries causing these errors. This can really help identify any unnecessary GPU-related parts. Please feel free to reach out if you’d like a hand tweaking the configuration or figuring out where the issue is coming from—I’m here to help!

We are seeing the same Any updates?

If you're still seeing those errors, it seems like the LiveKit agent server is trying to start GPU-related features even though it's running in a CPU-only setup. No worries—let’s work through this together: • Disable GPU Features: Take a look at your LiveKit agent server’s configuration for any settings or environment variables tied to GPU acceleration (like CUDA or VAAPI). Make sure these are disabled or set to CPU-only mode. • Use a CPU-Compatible Image: Double-check that the Docker image you’re using is meant for CPU-only environments and doesn’t have GPU dependencies. • Inspect Logs: Have a peek at the container logs to spot which modules or libraries might be causing these errors. This can help us figure out if there are any unnecessary GPU-related dependencies to remove. If you’ve already given these a try and the problem is still there, please let me know. I’m here to help and we can dig deeper into troubleshooting together!

No! @aloof-table-88981

I don't have the answer you're looking for. You could also try asking your question: • in one of the other Slack channels or • to https://deepwiki.com/livekit/livekit_composite which is trained on all LiveKit source code If you find the answer, please post it here to help others!

@fierce-dog-7777 we have our agents running daily since last couple of months This issue started happening since today

Same here @aloof-table-88981

It is safe to ignore those CUDA messages. Will drop the log levels on them so they are annoying