Save Bytes Bridge (Yogurt Nodes)

ComfyUI has a save node for images, one for masks, even one for text. Raw bytes - the thing you get from API downloads, serialized pickles, audio blobs - have nowhere to go by default. YogurtSaveBytesBridge is that missing sink: it takes a BYTES value and writes it to disk where you can actually find it.

The inputs, in the order you'll actually touch them:

• bytes_data - the bytes to save.
• filename_prefix - the file's name prefix (default "ComfyUI"). It supports formatting tokens like %date:yyyy-MM-dd%, so you can get date-stamped filenames for free.
• output_dir - where to write, relative to ComfyUI's output folder. Leave blank for the default output directory.
• overwrite - "true" or "false". Whether an existing file at the target name gets replaced.
• suffix - the extension: .pkl, .dat, .bin, or "Custom".
• custom_suffix - if you pick "Custom", the extension you want here (and this field wins over suffix when it's set).

Outputs: the original bytes_data passes through, plus path - the absolute path of the file you just wrote, which is the useful one to wire elsewhere.

What it's actually for

Think of the pack's own serialization flow: YogurtSerializeAny pickles a Python object into bytes, and this node saves those bytes as a .pkl. Later, YogurtDeserializeAny reads it back. That's a full save/restore path for arbitrary graph state. The same node saves files pulled down by the pack's API image nodes, or any binary blob a node hands you.

The path output matters because downstream steps often want the location, not the data - feed it into a text log, an "upload this file" step, or just a YogurtSaveTextBridge to record where things went.

Install

Ships in ComfyUI-YogurtNodes:

cd ComfyUI/custom_nodes
git clone https://github.com/yogurt7771/ComfyUI-YogurtNodes.git
cd ComfyUI-YogurtNodes
pip install -r requirements.txt

or install ComfyUI-YogurtNodes via ComfyUI Manager and restart. It's under "Yogurt Nodes"; the display name includes the "(Yogurt Nodes)" suffix. No models, no API keys.

The only real gotcha is overwrite. Batch-generate files with the same prefix and overwrite off, and later runs will have to make the filename unique instead of clobbering the earlier files - which is usually what you want, but it means the path output isn't a stable constant across runs. If you're writing a known filename for a downstream consumer, set overwrite to true and accept the collision risk.