How aws-deadline/deadline-cloud-for-keyshot works
Main System Components
• Submitter Component: Represented by the KeyShot submitter source code (located in src/deadline/keyshot_submitter) and its associated installation files. This component is responsible for running inside KeyShot’s Scripting Console, extracting scene info, packaging it appropriately (including converting file paths for attachments and generating a job bundle), and then initiating the job submission.
• Adaptor Component: Found in src/deadline/keyshot_adaptor along with its client counterpart (KeyShotClient). It implements the OpenJD adaptor interface, launching KeyShot on the worker hosts and handling the “sticky” session for rendering.
• AWS Deadline Cloud Interaction: Although the package itself isn’t implementing AWS services, the submitter and adaptor work together to create and submit a Job Bundle to the Deadline Cloud. Represent this external dependency with an “AWS Deadline Cloud Service” component.
• Testing Components: The “test” directory is organized into unit tests, integration tests, and installer tests. Although not a part of the runtime architecture, these components assure quality and could be mentioned as ensuring the reliability of the core system.
Relationships and Interactions
• The KeyShot Submitter (installed as a script in KeyShot) interacts with the local KeyShot application and collects the scene file and external references.
• After processing the scene, the submitter uses logic in the adaptor (or communicates with a client module) to generate an Open Job Description template and job bundle.
• The job bundle is then submitted through AWS Deadline Cloud’s client interface (using the Deadline client library) to the rendering farm (which may be service-managed or customer-managed).
• On the worker side, the KeyShot Adaptor is installed. Its job is to read the job payload, start KeyShot (or use keyshot_headless.exe), and process the render job.
• There is a clear directional data flow: KeyShot Interface (Submitter) → Job Bundle Generation → AWS Deadline Cloud (job submission) → Worker hosts (with Adaptor) → KeyShot rendering.
Architectural Patterns and Design Principles
• Modularity: The system is split into distinct modules (submitter, adaptor, client) to separate concerns.
• Extensibility: Leveraging the OpenJD adaptor runtime means the system is designed to support adaptation for standard job descriptions.
• Automation: CI/CD pipelines (via GitHub Actions and pipeline scripts) automate the build, test, and release process, which can be depicted as a supporting environment.
• Cross-Platform Elements: Although the primary job submission and rendering are handled differently on Windows and macOS (as mentioned in the README), focus on the overall integration rather than OS-specific details.
Relevant Technologies and Frameworks
• Python 3.9+, used throughout the project.
• KeyShot integration, leveraging its scripting interface.
• AWS Deadline Cloud for submission and management of rendering jobs.
• OpenJD Adaptor Runtime for Python.
• GitHub Actions, pipeline scripts, and CI/CD related tools for quality and release management.
• Module has a clear label.
• Optionally, include small notes next to each component indicating key technologies (e.g., “Python 3.9”, “OpenJD Adaptor”, “Deadline Cloud Client”) for clarity.