Data Types

Understanding the underlying data structure is crucial for seamless node interaction.

Frame-Based Data Format

Blobit uses a Frame-Based Data Format. This means almost all data passed between nodes is wrapped in a list, where each item represents a “frame”.

• Concept: [Frame1, Frame2, Frame3]
• Single Item: [Item] (Even a single image is a list of one image)
• Empty: []

Your backend BaseNode.normalize_to_frames() handles this automatically for inputs. Your BaseNode.pack_outputs() handles this for outputs.

Supported I/O Types

Blobit nodes communicate by passing file paths (Relative Strings) between the backend and frontend.

Type	Expected Python Return	Note
image	"/temp/filename.png"	Frontend expects an image path.
video	"/temp/filename.mp4"	Frontend expects a video path.
audio	"/temp/filename.wav"	Frontend expects an audio path.
string	"Any text"	Standard string.
number	1.0	Int or Float.
array	[...]	Generic list.

The “Path” Rule

Most BaseNodes exchange Paths (/temp/img.png), not heavy Tensors.

• Backend: Saves processed data to a temp file -> Returns relative path.
• Frontend: Receives path -> Normalizes to http://localhost... -> Displays it.

# Correct Return Pattern
return self.pack_outputs([saved_path_1, saved_path_2])
# Internal structure: {"output": [["/temp/img1.png"], ["/temp/img2.png"]]}