TT Context

TT Context is the pack's flexible junction box for its context system. If you already have a context, it passes it through while letting you drop new state in. If you don't, it can create one from whatever you feed it. Everything about it is optional - which is both the point and the trap.

It's the TenserTensor answer to the problem rgthree's Context nodes solve: your graph has too many wires and you want to bundle a bundle of state into a single object that travels down one connection. Where TT Base Context insists on the four core components up front, TT Context is the free-form version: plug in a context, a model, a CLIP, conditioning, a seed - whatever you happen to have - and it merges it all.

What's optional (everything)

The inputs, all optional: context, workflow_config, model, clip, vae, positive, negative, latent, image, seed, steps, cfg, sampler_name, scheduler. Feed it an existing context plus a new seed, and it updates just the seed. Feed it nothing but a model and a VAE, and it builds a minimal context from those.

The outputs mirror the inputs: CONTEXT plus individual sockets for WORKFLOW_CONFIG, MODEL, CLIP, VAE, POSITIVE, NEGATIVE, LATENT, IMAGE, SEED, STEPS, CFG, SAMPLER_NAME, SCHEDULER. So it also works as a breakout - a way to pull individual values back out of a context if something downstream wants a plain wire.

When to reach for it

This is the node you reach for mid-graph when you want to merge in a piece of state. Example: you're in a context flow, a ControlNet branch just produced a conditioning, and you want that in the context without rebuilding everything - TT Context takes your context and the new positive/negative, and the updated bundle continues. It's also the friendliest entry point if you want to try the context pattern without committing to TT Base Context's required-input discipline.

The honest warning: with all inputs optional, it's easy to build a context that's silently missing something, and downstream nodes will error at execution with "X is required" - the fix is tracing which optional field you forgot to feed. Also, this is a proprietary TT_CONTEXT socket, so it only talks to other TenserTensor context nodes. The breakout outputs (the plain MODEL, LATENT, etc. sockets) are standard types and portable; the context itself is not.

Install and the V1 note

cd ComfyUI/custom_nodes
git clone https://github.com/tenser-tensor/ComfyUI-TenserTensor

or ComfyUI Manager → "TenserTensor" → install → restart. Deps: gguf, kornia.

Standard for this pack: it's a legacy V1 node, parked in Deprecated/ after the migration to ComfyUI's API V3, with removal planned for a future major release. It works today and it's arguably the most useful of the pack's V1 context nodes for exploring the pattern - just don't build a workflow you'll rely on for years without noting the deprecation.