Inspect¶

Inspects any value flowing through your workflow. It logs a readable summary, shows the value as text on the node UI, and attempts to produce an image preview when possible (e.g., tensors or base64-encoded images).

Usage¶

Use this node to debug and understand intermediate data. Place it after any node whose output you want to inspect. It will format text/JSON for readability, log details (optional), and generate an image preview from tensors or base64 image data.

Inputs¶

Field	Required	Type	Description	Example
value	True	WILDCARD	The value to inspect. Accepts any type, including strings, JSON-like data, tensors, or objects that stringify. If the value is a tensor image or contains base64-encoded image data, a preview image will be generated.	{"image":"data:image/png;base64,iVBOR..."}
write_logs	True	BOOLEAN	When true, writes a formatted inspection block to the logs. Disable to avoid console/log clutter for large or frequent outputs.	true

Outputs¶

Field	Type	Description	Example
value	STRING	A formatted string representation of the inspected input. JSON is preserved and pretty-printed when detected.	{ "a": 1, "b": "text" }
image_preview	IMAGE	An image tensor preview when available. Derived from input tensors or base64-encoded image data; otherwise a blank placeholder image.	Tensor shaped like [1, H, W, C] with C in {1,3,4}

Important Notes¶

Formatting behavior: The output 'value' is always a STRING. Non-string inputs are serialized; strings that contain valid JSON are parsed and re-serialized for readability, while regular strings are passed through unchanged.
Image preview detection: The node attempts to render an image from: (a) image-like tensors, or (b) base64-encoded images found directly in the input string or inside JSON (e.g., data:image/... or raw base64 beginning with /9j/ for JPEG or iVBOR for PNG).
Tensor handling: If the input is a tensor shaped like an image, it will be used directly as the preview; otherwise a default blank preview is returned.
Logging control: Set write_logs=false to avoid writing to logs, useful for high-frequency or large payloads.
UI display: The node shows the inspected text in an on-node read-only text area to facilitate quick review.
Category: SALT/Utility

Troubleshooting¶

Blank image preview: Ensure the input is an image-like tensor ([B,H,W,C] or [H,W,C]) or includes a valid base64 image string (e.g., starting with data:image,..., /9j/ for JPEG, or iVBOR for PNG).
Unexpected stringification: The 'value' output is a string for inspection, not the original typed object. If you need to pass the original type forward, branch your workflow before Inspect.
Logs missing: Set write_logs=true to enable logging. Verify logging configuration if nothing appears.
Unreadable or compact JSON: Provide a valid JSON string or an object that serializes to JSON; invalid JSON will be treated as a plain string.
Performance with large data: For very large tensors or long strings, disable write_logs to reduce log overhead; previews fall back to a small blank image when decoding fails.