Troubleshooting

CUDA Toolkit Not Found

Problem: RightNow AI can't detect your CUDA installation.

Solutions:

1. Verify CUDA Toolkit 11.0-12.5 is installed (not just runtime)
2. Check installation path:
   • Windows: C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.x
   • Linux: /usr/local/cuda or /opt/cuda
3. Add CUDA bin directory to system PATH
4. Restart RightNow AI after installation

NCU Profiling Permission Denied

# Run Command Prompt as Administrator
reg add "HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\PerfProc\Performance" /v "Disable Performance Counters" /t REG_DWORD /d 0 /f

sudo usermod -a -G video $USER
sudo usermod -a -G render $USER
sudo usermod -a -G perf_users $USER
sudo reboot

sudo nvidia-smi -pm 1
sudo nvidia-smi -lgc 0,1980

OpenRouter API Key Not Working

Problem: "Invalid API key" or connection errors.

Solutions:

1. Verify API key at openrouter.ai/settings/keys
2. Check for extra spaces or special characters
3. Ensure account has credits/payment method
4. Test connection in Settings → AI Providers

Local Models Not Responding

Problem: Ollama/vLLM/LM Studio not working.

Solutions:

1. Ollama: Verify service is running: ollama list
2. vLLM: Check server endpoint is accessible
3. LM Studio: Ensure local server is started and model loaded
4. Verify firewall isn't blocking local connections

Kernels Not Profiling

Problem: Play buttons don't work or no metrics shown.

Solutions:

1. Ensure kernel contains __global__ functions
2. Check CUDA code compiles successfully with nvcc
3. Verify GPU supports compute capability 3.5+
4. Close other GPU-intensive applications
5. Try profiling a simple "Hello World" kernel first

Inaccurate Performance Metrics

Problem: Profiling results seem wrong or inconsistent.

Solutions:

1. Warm up GPU with multiple kernel runs
2. Use release build configuration (not debug)
3. Ensure sufficient GPU memory available
4. Profile on dedicated GPU (not integrated graphics)
5. Check for thermal throttling: nvidia-smi

Slow AI Responses

Problem: AI completions or chat responses are slow.

Solutions:

1. BYOK Users: Check OpenRouter account limits and credits
2. Local Models: Ensure sufficient RAM/VRAM for model size
3. Network: Verify stable internet connection
4. Try switching to a smaller/faster model temporarily

High Memory Usage

Problem: RightNow AI consuming too much RAM.

Solutions:

1. Close unused profiling panels
2. Limit chat history length in settings
3. Reduce autocomplete context window size
4. Restart RightNow AI periodically for long sessions