deepstream-dev

You can build an NVIDIA DeepStream pipeline using the pyservicemaker API, which automates pipeline construction and replaces verbose GStreamer boilerplate with a higher-level, Pythonic interface for real-time video analytics workflows.

Yes, dynamic source management is supported through REST-managed sources. This allows you to dynamically add or remove RTSP feeds and local videos from a running DeepStream pipeline without restarting the application.

The Pipeline API provides fine-grained control over DeepStream pipeline assembly, while the Flow API enables declarative pipeline construction. Both APIs support multi-stream batching, inference, tracking, and messaging components for video analytics.

You can configure nvinfer for primary inference using YAML or INI configuration files. The pyservicemaker API assembles the DeepStream pipeline to read these configs, performing inference on local videos or RTSP feeds with optional tracking and messaging.

Yes, pyservicemaker supports multi-stream batching across single or multi-stream inputs and includes telemetry probes for observability. This enables production-grade DeepStream deployments with real-time video analytics and performance monitoring.

Use pyservicemaker when you need to accelerate DeepStream development and governance, avoid intricate GStreamer pipeline orchestration boilerplate, and require features like dynamic REST-managed sources, multi-stream batching, and telemetry probes.