Griptape RAG Response: Text Chunks Module

Every other response module in this pack generates an answer. This one doesn't - it just hands back the retrieved text chunks, untouched, as the response. No LLM call, no summary, no opinion. That sounds useless until you think about what it enables: you get the retrieval results out of the RAG pipeline as plain data, to pipe into something else. Maybe you want to feed raw chunks into your own prompt elsewhere, build a summarizer with different rules, or just inspect exactly what retrieval found before committing to an answer.

It's the "let me see the receipts" node, and it's also the zero-token response module: because nothing generates text, it costs nothing per call beyond retrieval.

The inputs that matter

None. This is the only RAG module in the set with an empty optional-input list - no prompt_driver, no rulesets. It's deliberately a passthrough. The one thing you might expect but won't find here is any control over how many chunks come back; that's governed by the retrieval stage and the vector store / text loader modules upstream, not by this node.

How it works

Under the hood it constructs a Griptape TextChunksResponseRagModule. In the response stage, instead of prompting an LLM with the retrieved text, it returns the text chunks themselves as the stage's result. Everything downstream - whatever consumes the RAG engine's response - sees the raw material rather than a generated answer.

What it outputs

A MODULE_LIST, wired into response_stage_modules on the Griptape RAG Engine. If your pipeline's goal is "give me the relevant passages," this is the terminal module; if the goal is "answer my question," use the Prompt or Footnote response modules instead. It's a fork in the road, not a replacement.

Installing it

Ships in the pack - ComfyUI Manager → Griptape → ComfyUI-Griptape, restart, or

cd ComfyUI/custom_nodes
git clone https://github.com/griptape-ai/ComfyUI-Griptape

Common issues

The main confusion is expecting an answer and getting chunks - if you wired this in expecting "the RAG module" to behave like the prompt one, the output format will look wrong. That's not a bug; it's the module doing exactly what its name says. Also remember it's still a response-stage module: it needs retrieval upstream to have anything to pass through, and it needs a consumer downstream that can handle raw chunk text rather than prose. Pack-wide torch/version issues apply as usual - reinstall torch with the cu121 wheels if ComfyUI breaks, python -m pip install griptape -U for stale-library import errors.