Best practice for integrating an existing Nuke plate directory structure with Flow Production Tracking Toolkit?

Flow Production Tracking
, , ,
1

Hi everyone,

I’m setting up a new Flow Production Tracking (ShotGrid Toolkit) pipeline for a small VFX studio (~30 artists) using Nuke on Rocky Linux 9, and I’d like to follow Autodesk best practices rather than reinvent the wheel.

Our existing production data already lives on disk in the following structure:



/home/mayurlagad/Projects/
└── BT/
    └── Plates/
        ├── REEL 01/
        │   ├── BT_R1_SH001/
        │   │   ├── BT_R1_SH001.0000048.exr
        │   │   ├── BT_R1_SH001.0000049.exr
        │   │   └── ...
        │   ├── BT_R1_SH002/
        │   └── ...
        ├── REEL 04/
        └── REEL 05/

Our mapping is:

  • Project = BT
  • Sequence = REEL 01
  • Shot = BT_R1_SH001
  • Plates already exist on disk before artists start work.

What I’d like to achieve is a workflow similar to what larger studios use:

  1. Scan the existing Plates directory.
  2. Automatically create missing Sequences and Shots in Flow Production Tracking.
  3. Automatically associate each EXR sequence with the corresponding Shot.
  4. Artist opens Flow Production Tracking Desktop.
  5. Selects Project → Sequence → Shot → Comp Task.
  6. Launches Nuke.
  7. Nuke automatically:
    *
    opens/creates the correct work file,
    *
    loads the correct plate as a Read node,
    *
    sets frame range,
    *
    sets project format,
    *
    creates the standard Write nodes.

I have already explored the default Toolkit configuration (tk-config-default2) and understand how tk-multi-workfiles2, tk-multi-loader2, and tk-multi-publish2 work.

My questions are:

  1. How do most studios handle existing plate directories like this?
  2. Do studios typically:
    *
    register plates as PublishedFile entities,
    *
    store plate paths in custom Shot fields,
    *
    extend templates.yml,
    *
    or use another approach?
  3. Is there a recommended Autodesk Toolkit pattern for automatically loading a plate into Nuke on launch, or is this generally implemented through custom hooks?
  4. Where is the best place for the automation that scans the filesystem and creates Sequences/Shots/PublishedFiles?
    *
    Standalone Python script?
    *
    Event Daemon?
    *
    Toolkit command?
    *
    Something else?
  5. If you were designing a modern Toolkit pipeline from scratch today, would you keep the existing plate directory structure or reorganize it to match Toolkit’s default schema?

I’m specifically looking for best practices used in production studios, not just something that works.

Any architecture recommendations, examples, or open-source implementations would be greatly appreciated.

Thanks!