Sampler → Dict🗂️

The advanced sampling API treats the sampler - the solver you pick, like Euler or DPM++ - as an object you hand to a sampler node, not just a dropdown inside it. DictAddSampler stores that object in your dict under a named key. If you're assembling sampling from explicit parts, this is how the sampler itself rides the bus.

It's grouped under Add/Advanced with the guider, noise, and sigmas nodes. The SAMPLER value is a wired input from a sampler-selecting node (think KSamplerSelect), validated against the sampler type before storage.

How it works

Same Dict Tools engine as the whole pack: validate, store under the key, return a new frozendict without mutating your input, and sort keys alphabetically on update. The sampler is held by reference, so bundling is cheap.

The pattern: advanced sampling needs guider + noise + sampler + sigmas to converge on one node. The four Dict Add nodes let you pack them all into one dict, drag a single connection anywhere, and extract the pieces at the sampler. For multi-stage workflows where two stages use different solvers, distinct keys mean each stage pulls its own sampler from the same dict.

The inputs that matter

• key (STRING) - the name you'll fetch the sampler back by.
• cleanup_key (BOOL, default on) - trims spaces/newlines from the key.
• dict (DICT, optional) - an existing dict to extend, or empty to start one.
• value (SAMPLER, optional) - the wired sampler object.

Output is dict (DICT). Get it back with DictExtractSampler.

Install

ComfyUI Manager: search Dict Tools, install, restart. Or:

cd ComfyUI/custom_nodes
git clone https://github.com/Lex-DRL/ComfyUI-DictTools

Restart after. Only dependency is frozendict, auto-installed by Manager.

Gotchas

Pack-wide: requires ComfyUI 0.18.0+ (newer node API). The Advanced nodes are doubly sensitive to version: if the SAMPLER input type doesn't appear at all, your ComfyUI predates the explicit-sampler API - update it.

And the usual snapshot reminder: picking a different solver after you've added the sampler to the dict doesn't change what's stored. Re-add the new sampler object if you switch mid-workflow, or you'll keep sampling with the old solver and spend a while wondering why nothing changed.