Griptape Run: Text Summary

Summarize a chunk of text with an LLM, in one node. That's it - that's the pitch. Griptape Run: Text Summary is the simplest useful text node in this pack, and it's the one you'll wire in whenever something upstream produces a wall of text and something downstream wants the short version.

It earns its place because summarization is the connective tissue of agent pipelines. An image-description node gives you three paragraphs about a photo; a text-summary node condenses it to one line that fits in a final prompt. An audio transcription produces forty minutes of rambling; this turns it into talking points. The pattern - long thing in, tight thing out - shows up everywhere, and having it as a first-class node beats doing it with a raw prompt on Run Agent.

Under the hood it's a TextSummaryTask, Griptape's dedicated summarization task - the framework's own machinery for condensing text, not just a reworded prompt. The summary is produced by your connected agent's prompt driver, so model choice is yours. It supports key_value_replacement for {{ key }} template fills like the rest of the pack.

The inputs

• STRING - the text to summarize (required).
• agent - optional; the agent whose model does the summarizing.
• input_string / key_value_replacement - appendix and template fill.

Outputs: OUTPUT (the summary) and AGENT.

Installing

ComfyUI Manager → search "Griptape" → Install, or:

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

Restart ComfyUI. Deps: griptape[all], python-dotenv. For local-only setups, an Ollama-based agent config keeps this completely offline.

Common gotchas

Summary quality is model quality - a weak local model will faithfully reproduce the first sentence and call it a summary. If you're getting lazy summaries, the model is the bottleneck, not the node.

The pack-wide environmental issues still apply: torch version conflicts on Nvidia (reinstall with --extra-index-url https://download.pytorch.org/whl/cu121) and the stale-griptape ImportError: cannot import name 'OllamaPromptDriver' (fix: python -m pip install griptape -U). And if you're in a long workflow, remember each summary is a fresh model call - chain a few and you're spending like it's a video card.