📦 RH Param Bundle

RH Param Bundle is pure glue: it takes up to ten separate RH_PARAMS chains and bundles them into a single RH_PARAM_BUNDLE - the pack's "list of jobs" type. It does no uploading, no executing, no math. Its entire job is to arrange your parameter sets so RH Batch Execute can turn each one into its own cloud task.

Think of it as the batch counterpart to RH Param. A single RH Param chain describes one run: "set node 3.text to this, node 5.seed to that." A Param Bundle says "here are five different descriptions, run the workflow for each." Each input you connect is one run; the more inputs you fill, the more tasks Batch Execute will fire.

Inputs

Ten optional inputs, params_1 through params_10, each accepting an RH_PARAMS chain. That's the whole node. Connect one chain and the bundle has one job; connect eight and it has eight. The inputs are optional because a bundle with no connected inputs is valid-ish - though the node warns about it and hands back an empty list.

How it works

The bundle() method collects every non-None input, wraps each in its own job, and returns the list. Each job stays independent - a parameter chain with {nodeId, fieldName, fieldValue} entries - so when Batch Execute iterates the bundle it submits one task per job with that job's params. The order of your jobs matches the input order (params_1 first), which is handy when you're pairing jobs with results later.

Where it fits

The natural assembly line is:

1. Build N distinct parameter chains (N separate RH Param chains, or RH Batch Upload Image / RH Multi-Input Image outputs).
2. Feed them into this node's params_1..N.
3. Feed the bundle into RH Batch Execute along with your config.
4. Collect the comma-separated task IDs and hand them to RH Download.

That's the full batch pipeline the README's batch_processing.json example walks through.

Gotchas

Each connected input becomes a separate cloud task with its own credits. Ten inputs, ten billed runs - easy to forget until the bill surprises you. Also, a bundle only stores jobs; nothing runs until Batch Execute fires. And because it's dumb glue, it won't dedupe or validate your chains - if a job has a wrong nodeId, the upload/execute side will happily submit it and the cloud will error or ignore it. Garbage in, billable garbage out.

Install is pack-wide: ComfyUI Manager (search ComfyUI_RH_API) or git clone https://github.com/xuchenxu168/ComfyUI_RH_API.git into custom_nodes + pip install -r requirements.txt, restart.