How erofcon/rust-vision works
Project type and purpose
• This is a distributed, real-time video-processing pipeline built in Rust.
• It ingests video streams via GStreamer, runs object-detection and tracking (using ONNX Runtime), publishes tasks over RabbitMQ, processes them in worker pools, stores results in a database/S3, and exposes an HTTP API (and CLI) to query processed data.
File structure
• Top-level “crates/” folder holds reusable Rust libraries:
• common (shared data types, config, utilities)
• queue (RabbitMQ producer/consumer)
• gst-streaming & streaming (GStreamer-based ingest pipelines)
• detection, motion, inference (analytics libraries)
• storage (DB/S3 repository layer)
• api (HTTP server, handlers)
• “app/bin” and “cli/” are executables: the long-running worker service and command-line client.
• “config/” contains environment-specific TOML configs.
• “docker/” and docker-compose.yml orchestrate API, worker, RabbitMQ, DB, Nginx reverse-proxy.
• SQL migration scripts under root and migrations/ initialize PostgreSQL schemas.
README
• Describes real-time video capture, analytics, and tracking.
• Mentions GStreamer for capture, ORT for inference.
• Points to docs/api.md for API details.
• Warns project is under active development.
Detailing essentials
• Microservice or crate as its own component block.
• Environment configs feeding into services.
• External dependencies (e.g., camera feeds, S3).
• How Docker-Compose wires containers together on a virtual network.
• Protocols (AMQP, HTTP, SQL, gRPC if any).