What problem does it solve? Building a GPU-accelerated SOP compliance microservice involves coordinating DeepStream pipelines, Triton GEBD models, embedded vLLM inference, SSE streaming, Kafka output, and Basler camera support — with dozens of subtle ordering and configuration rules that are easy to get wrong. This Skill encodes the complete architecture, critical rules, and verification workflow so an AI assistant can generate, deploy, evaluate, and debug the service correctly. ## Core Features & Use Cases - Full microservice generation: Produces the FastAPI server, SOPProcessManager wiring, DeepStream pyservicemaker pipeline with DDM event-boundary detection, VLM classification via Cosmos Reason, SOP sequence checker, and Docker build files. - Evaluation and debugging workflow: Guides static validation, Docker build, service launch, API test suites, curl verification, and latency measurement (TTFC, C2C, camera chunk_e2e). - Critical rule enforcement: Encodes hard-won constraints such as VLM warmup before DDM, stream:true for live inputs, named-kwargs factories, and TensorRT stream synchronization. - Use Case: A factory engineer asks the assistant to verify whether operators perform assembly steps in order on a Basler camera feed; the Skill produces the correct chat-completions request and the underlying service configuration. ## Quick Start Ask the assistant to use the deepstream-sop skill to generate the SOP inference microservice scaffold following the example prompt in its references directory.
