[Legacy] H3 Continuum Save Session

H3 Continuum Save Session is the disk-sink for the V2-era continuation workflow. Its job is narrow and it does it well: take the H3_CONTINUUM_SESSION object that a H3 Continuum Sampler V2 produces, and atomically save every accepted chunk latent plus the session metadata to disk. The word "accepted" matters - the session only persists chunks the sampler actually accepted, so a partially completed run saves the good prefix rather than pretending it finished.

Inputs: session (the session handle), prefix (default h3_continuum_session), and slot (1, up to 9999). Outputs: session (passed through) and path (a STRING giving you the file location - connect it to a Preview as Text if you want to see where things landed). It's marked as an output node, meaning it can terminate a graph on its own.

The point of all this: later, a Load Session node with the same prefix and slot pulls the saved chunks back, and the sampler's reroll_from_chunk reuses everything before N and regenerates N onward. You don't re-run the whole 15-second chain to fix a bad ending - you reuse the saved prefix.

Install

cd ComfyUI/custom_nodes
git clone https://github.com/ukr8b3g-cmyk/ComfyUI-H3-Continuum.git

or via ComfyUI Manager ("H3 Continuum"), then restart. No extra pip packages; needs a MiniMax H3 ComfyUI environment (weights not bundled) and ComfyUI v0.32.0+.

Before you go building session plumbing into a fresh workflow, know the modern answer: the V3.2.4+ samplers have run_storage = Save + Auto Resume built in, which atomically saves raw AV chunks and resumes them automatically - no Save/Load session nodes, no manual slot bookkeeping. Save Session is the node you'll meet in V2-era workflows and the occasional tutorial that explains session-based reroll by hand. It still works; it's just more manual than the thing that replaced it. Slot numbering is your memory, so pick stable prefix/slot values if you rely on this across runs.