# A Tour of Orchestrations Walk the Orchestrations list, open one in the Builder, and see how nodes, ports and Roles fit together. - Page: https://perceivo.ai/university/orch-02-tour-of-orchestrations - JSON: https://perceivo.ai/registry/film/orch-02-tour-of-orchestrations.json - Kind: film - Revision: 2 (published 2026-10-18, produced 2026-10-15) - Previous revision 1 (published 2026-10-07): https://perceivo.ai/registry/film/orch-02-tour-of-orchestrations/r1.json - SDK version: none ## Validation - SDK: not applicable (UI walkthrough; no component or orchestration is offered for download) - Platform: not applicable (UI walkthrough; no component or orchestration is offered for download) ## Transcript [0:00] Introduction to Orchestrations, Episode 2: A Tour of Orchestrations [0:06] Welcome to our Introduction to Orchestrations series. [0:10] This video introduces the web-based interface for building, running and evaluating orchestrations: automated pipelines designed to achieve complex, multi-step goals. [0:22] Orchestrations are part of the Studio: a collection of tools you use to build AI solutions. [0:29] Orchestrations let you combine components into intelligent workflows. [0:35] Orchestrations you import or build are listed here. [0:39] In the next episode we show you how to import one. [0:42] For now, let's open one we have already built. [0:47] Each orchestration has five tabs. [0:49] The Builder, which you are looking at now, is where you develop and validate your orchestration. [0:55] We'll come back here in a minute. [0:58] You use the Runner to start runs and watch them live. [1:02] Observe is where you explore execution history and results, including outcomes, metrics, logs and costs. [1:11] Roles gives you fine-grained control over who can see, edit and run this orchestration. [1:18] And Settings: connection details for the Perceivo API. [1:22] Expand any call to see its address, the response it returns, and a ready-made cURL command. [1:29] Under Configuration, Actor Resources sets the compute and memory allocated to this orchestration. [1:36] And Versions is your version history, where you can reopen any earlier version. [1:42] Back to the Builder. [1:44] There are three main areas here: the palette on the left, the canvas in the middle, and the inspector on the right. [1:52] The palette is your parts bin. [1:54] At the top are the primitives, the building blocks: start and end, agents, tools, state machines, policies, review steps and adapters. [2:05] Below them is the quick select. [2:07] It lists the parts you use, most frequent first. [2:11] Favorite one, and it is pinned to the top, so it is always within reach. [2:17] The Gallery shows everything you can use. [2:20] View all your components, or filter on just templates, orchestrations, or components. [2:27] We cover nested orchestrations and templates in episode six. [2:32] The canvas is where you construct orchestrations by linking nodes with edges. [2:37] You drag parts onto the canvas, connect their ports, and configure them in the inspector. [2:43] In this simple orchestration, execution flows one way, from start to finish. [2:49] We call this path from start to end the main flow. [2:52] All ports on the main flow are one-way. [2:55] Other connections are two-way: they send a request and receive a response. [3:02] In Perceivo, both the request and the response are typed, so a component can only connect to compatible components. [3:09] Episode four is all about those connection rules. [3:14] Now let's run it. [3:15] The Start component has a special external input port. [3:19] When you click Start Run, it reads that configuration and opens a dialog that prompts for inputs, based on the contract defined on the orchestration's Start component. [3:29] That is how inputs reach your orchestration: artifacts, metadata, or both. [3:35] Here we are not passing an artifact. [3:38] We are providing a request data object that has a prompt, which is a string, and a maximum length, which is an integer. [3:46] Start Run queues a job, sized to the compute and memory you set in Settings. [3:51] The job ends as soon as the run completes. [3:55] Run one-shot like this, you pay for compute only while the run's pod is allocated. [4:01] It is a true on-demand pricing model. [4:05] Your agent can also start the run with a post to the API, or with the Perceivo client SDK. [4:11] Same result. [4:13] Executing components light up while they are active. [4:16] When each one completes, its latency and cost appear above the node on the canvas, and are tallied in the inspector on the right. [4:26] Done in seconds. [4:28] Latency and cost, logged for every component. [4:37] Everything in this series is open source. [4:40] You'll find this orchestration, and every example that follows, in the Perceivo SDK repository under examples. [4:48] The best way to learn this platform is to build with it. [4:53] Next time, we show you how to import and export orchestrations, so you can bring every example in this series into your own workspace. [5:01] See you there.