LM Studio Sequential Prompt

LM Studio Sequential Prompt is the plain version of the pack's chain node: it runs a list of prompts one at a time against an LM Studio model, each step feeding the last, and hands you the final answer. No system prompt, no top_p/top_k, no token cap - just the chain, temperature, and seed. If you've looked at Sequential Prompt Advanced and thought "I don't need half those knobs," this is the one that skips the clutter.

How it works

Everything in the prompt_templates field gets split on the delimiter (default ---), and each chunk becomes one step. With send_as_chat off, every step sends the full accumulated conversation, so the model always has context from earlier steps - that's the whole point of a chain. Flip it on and each step goes out alone, which is faster and makes the steps independent. temperature and seed are passed through to the model on every step; filter_thinking (on by default) strips <think>/<thinking> reasoning tags out of the replies, which matters for reasoning models.

The one input to be careful with is the delimiter: pick a string that never appears inside your actual prompts, or a step gets split into two. And note it's a per-step cost model - a long chain means that many sequential requests to LM Studio, each one carrying the whole history unless send_as_chat is on.

Outputs

• final_response - the last step's answer. This is the one you wire into a text encoder or wherever the chain's result needs to go.
• previous_conv - the whole conversation as JSON, for threading into another LM Studio node.
• read_conv - the same history in human-readable form.

Installing

Same as every node in the pack: ComfyUI Manager → search "comfyui-lmstudio", or

cd ComfyUI/custom_nodes
git clone https://github.com/SiegeKeebsOffical/comfyui-lmstudio

then restart ComfyUI. Load a model in LM Studio and start its local server (default port 1234); no models download through the pack. If you hit ModuleNotFoundError: No module named 'httpx', pip install httpx and restart.

The honest take

This node and its Advanced sibling do the same thing; the difference is purely whether you want a system prompt and sampler controls on the chain. Start here if you're new to chaining - fewer inputs to trip over. One habit worth keeping either way: turn debug on for your first run so you can see each step's request printed to the console, then turn it off. It's the difference between guessing which step produced the garbage and knowing.