nvidia-nim-unified-skill

Use when a task needs NVIDIA NIM OCR, page elements, table structure, graphic element detection, passage reranking, or text embedding through one shared routing layer. Selects the right capability or multi-step workflow from a free-form query, prepares requests, and can invoke the configured endpoint. Also supports RAG pipeline mode: file/URL → OCR → semantic chunking → structured output.

Python backend worker for CLI tools and API servers