isabella232 / aip-orchestrator

A light-weight AIP simulator to test processors

AIP Orchestrator

This repo contains a light-weight simulator of the AI Inference Platform (AIP). It can send InferenceRequests to a processor at some frames-per-second, and print the InferenceResponses received.

The real AIP is Palantir Gotham’s platform to perform real-time augmentations and computations on input data. AIP is the base platform that accepts inputs and decodes them. Then, it runs the inputs through zero or more “processor” services, which are responsible for various sorts of processing. Please read the AIP SDK Documentation for more information.

The orchestrator comes with a sample image of dimensions 2048 x 2048 x 3 that it sends to the processor in either of the RGB888, BGR888, PNG or TIFF formats (depending on what the processor specifies in the configuration). It currently does not support using custom images, as the prime goal of the orchestrator is to test that a processor can establish basic communication with the AIP interface. To run your own images/video/inputs through, please use AIP.

Running the Orchestrator

Please refer to the QuickStart Guide for how to run this orchestrator along with a sample Python inference processor.