Project Director v1.0 - Akki
The project path node that stops Akki's film pipeline from dumping files everywhere
- project_path
If you've loaded an AkkiNodes workflow and wondered what the Project Director node actually does, here's the short version: it exists so every other node in the suite writes to the same folder. That's it. It's two text boxes and a string output, and it's the least glamorous node in the "Your Personal AI Film Studio" pack - but the whole file organization of the pipeline hangs off it.
The suite is built as a "Script to Screen" chain: story → screenplay → shot list CSV → lookdev assets → final per-shot prompts. Every stage produces text files and images that downstream nodes have to find again. Those loaders and savers (Scene Choreography Loader, Video Prompt Loader, Shot Asset Loader, Save Text File) all take a project_path. The Project Director is what generates that path consistently, so you type your project name once instead of pasting "output/AKKILLM/Project01" into a dozen nodes and praying they match.
How it works
Under the hood it's almost insultingly simple - and that's the point. You give it:
- project_name (default
Project01) - the folder name for this production. - base_output_folder (default
output/AKKILLM) - where the project folder lives, relative to ComfyUI's output directory.
The node sanitizes the project name (keeps only alphanumeric characters, spaces and underscores - so My:Film/2 becomes a valid folder name), then joins them into a single project_path string. No LLM, no API, no model download. It's a pure path builder.
One detail worth knowing: that path is relative to your ComfyUI output/ directory. The loaders in the pack prepend folder_paths.get_output_directory() themselves, so output/AKKILLM/Project01 resolves to ComfyUI/output/AKKILLM/Project01 on disk. Wire the project_path output straight into the loaders' project_path inputs and everything lines up - that's the whole reason the field exists.
Installing
The Project Director is part of the full AkkiNodes LLM Suite, so you install the whole pack:
cd ComfyUI/custom_nodes
git clone https://github.com/routhakash/AkkiNodes-LLM-Suite-for-ComfyUI
On Windows, right-click install.bat in the new folder and Run as administrator once - it activates your ComfyUI venv and installs llama-cpp-python==0.3.9 (with the cu122 wheel index) plus openai. Then drop a GGUF model into ComfyUI/models/llms/ and restart ComfyUI. On Linux/macOS there's no bat file, so run those two pip installs against your ComfyUI environment yourself. The other nodes in the pack need the LLM; this one doesn't, but it still only shows up once the pack loads.
Gotchas
- If you change the project name later, every downstream file moves to a new folder - old outputs stay where they were, and loaders that look in the old path will start returning
ERROR:strings. Pick the name before you generate a library of assets. - The output is a plain string. If you don't wire it in and just eyeball it in a Text node, you'll see the relative path, which looks wrong if you're used to absolute paths. It isn't.
- This pack is a solo, heavily-versioned effort with essentially no community discussion behind it - version numbers in node titles (here "v1.0") bump often, but this node's contract has been stable. Treat it as the stable anchor it is.
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| project_name | STRING | Project01 | — |
| base_output_folder | STRING | output/AKKILLM | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| project_path | STRING | — |