OpenOutpainter Serving Output Interrogate

The companion to the Input Interrogate node, and the piece that makes the whole interrogate round-trip land. openOutpaint asked your workflow to caption an image; your graph produced some text; this output node wraps it in the response the canvas expects and lets it show up in the prompt box.

It's a terminal output node with two required inputs and no outputs:

• oop_request - the request object, threaded through from the Serving node.
• caption (STRING) - the caption text your workflow generated. This is a forced input, so you wire it from whatever produced the text, not type it in.

How it works

Command-aware and lazy, in the same style as every node in this pack: it only does anything when the pending request is an interrogate call and caption is present. When it fires, it finalizes the request with:

{"caption": "<your text>"}

The HTTP handler waiting on the request wakes up, writes that JSON back to openOutpaint, and the canvas's interrogate UI displays the result - typically filling the prompt field so you can edit or reuse it. The comment in the source sums up the contract: the canvas sends an image with "model":"clip", and the response is just the caption string. Everything else is your workflow's business.

Wiring and the failure modes

Wire caption from your captioning/tagging node - the Input Interrogate node hands the image up, your captioner turns it into text, this node hands the text back. Get that chain right and the interrogate tool "just works" from the canvas. The most common mistake is feeding caption from the wrong place, or skipping the captioner entirely and shipping an empty string - the canvas will happily accept it, and you'll wonder why the prompt box is blank.

The pack's usual gotchas apply: an error upstream leaves the canvas request hanging until you re-run the workflow with the same request_id, and this pack runs one serving workflow at a time, so a second Serving node is unsupported territory. Install is the family standard - ComfyUI Manager (search "OpenOutpaint ComfyUI Interface") or git clone https://github.com/without-ordinary/openoutpaint_comfyui_interface into ComfyUI/custom_nodes, restart - with opencv-python as the only dependency worth noting and no model downloads involved.