Essential AI and profiling capabilities for CUDA development
Context-aware CUDA completions with Fill-in-the-Middle (FIM) optimization. Supports 20+ FIM-capable models including DeepSeek R1, Codestral 2501, and StarCoder2.
Select any code and press Ctrl+K to describe changes in natural language:
Full project context with CUDA-specific knowledge:
Production-grade profiling using nv-nsight-cu-cli with comprehensive hardware metrics:
Profile specific __global__ functions with targeted analysis
Full executable profiling with complete call graphs
Direct nv-nsight-cu-cli integration with custom metrics
🟢 Green
>80% efficiency (optimized kernels)
🟡 Orange
40-80% efficiency (moderate performance)
🔴 Red
<40% efficiency (needs optimization)
nvidia-smi integration for live metricsComplete offline capability with no data leaving your machine: