FAL Context Store

Of the four nodes in this pack, FAL Context Store is the one that makes people do a double take. One input, one output, and under the hood it returns exactly what you gave it. That's the whole node. It's a pass-through - a string goes in and the same string comes out - and the reason it exists is workflow plumbing, not computation.

Why a pass-through exists

FLUX Kontext workflows chain edits while keeping a subject's identity intact. FAL Text2Image emits a context_id_out - a request id - and the pack's intended flow is to carry that id into a later FAL Image2Image step so the edit chain keeps the same context across runs. That value has to live somewhere in your graph, and this node is where. Think of it as a reroute node for a string: it gives the wire a home, a name, and a place to branch, so you don't lose track of which context a later edit belongs to. The README's minimal Kontext workflow is literally T2I → Context Store → Image2Image.

You can absolutely skip it and wire context_id_out straight into the next node's context_id input. Nobody will stop you. It's organizational, not functional - the same role rgthree's reroute nodes play for the rest of your graph.

One honest caveat

The plumbing here is ahead of the request-side code. The nodes accept context_id on both generation nodes and emit context_id_out, but in the shipped code that value isn't actually injected into the fal API payload - and this node just echoes the string through. The README claims context-id reuse is "verified" and maintains identity across runs, but the source on disk doesn't send it yet. So treat Kontext identity-chaining as work-in-progress: the graph wiring is ready, the payload isn't. If you're on a newer commit than the one I checked, re-verify before you bet a workflow on it.

Inputs, outputs, and cost

It's the cleanest schema in the pack: context_id_in (STRING, required) in, context_id (STRING) out. That's it. And because it never touches the network, it's the one node in the pack that costs you nothing - no API call, no credits, no key. It works even if your FAL_KEY is missing.

Install

It ships in the ComfyUI_fal_image pack with the other three FAL nodes - install once, use all of them:

cd ComfyUI/custom_nodes
git clone https://github.com/Paddel87/ComfyUI_fal_image.git
pip install -r ComfyUI_fal_image/requirements.txt
export FAL_KEY="fal-..."

Restart ComfyUI and it appears under fal.ai/Image. ComfyUI Manager (search "ComfyUI_fal_image") works too; on the Windows portable build, use python_embeded/python.exe -m pip install ....

Common issues

The one real gotcha is expectation: don't expect this node to do anything on its own, and don't confuse it with storing images or caching results. It holds an id, not a render. If a Kontext workflow isn't holding identity across edits, the culprit is upstream - likely the context_id value not reaching the API in the current code, which this node can't fix. And if you're using a non-Kontext model like plain FLUX Dev, a context id is meaningless anyway; this node only earns its keep in Kontext edit chains.

It's a tiny utility in a small solo-authored pack, but it's the honest kind of utility: transparent about being a wire, useful for exactly one job.