FD Workflow Image Saver Advanced

Feidorian_WorkflowImageSaverAdvanced closes out the loop for the FD Workflow Image Loader Advanced. Feed it the images your pipeline produced, and it writes each one into the per-file subfolder the loader set up, under the product_ name the loader reserved. On the next run, the loader sees a product_ file with a matching processed_ file, considers the pair complete, and moves the product into APPROVED_IMAGES. That's the whole batch cycle, end to end.

How it works. Like the non-advanced saver, it doesn't ask you for a path or filename - it reads state that the Advanced loader set during its pass (file_name and file_dir tracking params). It writes the PNG, embeds prompt_text and, when with_workflow is on, the full workflow JSON as metadata, so the finished product is a self-contained file that carries its own caption and graph.

The difference from the plain saver is the to_cached boolean, and it's worth understanding because it changes where your file goes:

• to_cached off (default) - saves product_<name>.png into the working subfolder. That product gets approved into APPROVED_IMAGES on the next loader pass, and the batch's task counter advances.
• to_cached on - saves a timestamp-named file into CACHED_IMAGES instead. No product is created, nothing gets approved, and the batch counter doesn't advance.

So the cache path is your "preview without committing" lane: run the workflow, eyeball the cached result, and only turn caching off once you're happy. It's a surprisingly clean way to QA a batch before you burn through your source images for real.

Inputs. images (required), prompt_text, with_workflow (default true), and to_cached (default false). No outputs - it's the end of the line, an output node.

Common issues. The classic one is the same as its sibling: "Missing Path. Path must be provided to loader node." means the saver ran without the Advanced loader having set up state in the same session. Keep the pair in one graph. Second: to_cached left on by accident is easy to miss - your products silently accumulate in CACHED_IMAGES while the base folder keeps its full queue. Third, be aware the pack is archived, so this behavior is frozen; if the loader's strict "corrupted directory" rules annoy you, there's no fix coming.

Install. ComfyUI Manager (search "feidorian-ComfyNodes"), or:

cd ComfyUI/custom_nodes
git clone https://github.com/Feidorian/feidorian-ComfyNodes

Restart ComfyUI after. No requirements.txt, no models, nothing heavy - it's pure Python over ComfyUI's own image handling, which is exactly what an archived pack should be: easy to install, frozen in place, and still useful.