Griptape Agent Config: Ollama [DEPRECATED]

Here's the OG local config: the deprecated Ollama structure config was the pack's first way to run a Griptape agent against a local Ollama model, no cloud involved. It still works and legacy workflows reference it constantly, but it's been replaced by "Griptape Agent Config: Ollama Drivers," which bundles an embedding driver and a local vector store so agents actually get memory. If you're staring at this node in an old workflow, here's the deal.

What it is

A single config node that points a Griptape agent's prompt driver at Ollama. Three inputs are required: prompt_model (a free-text model name, e.g. llama3), base_url (default http://127.0.0.1), and port (default 11434 - Ollama's default). The optional side has the old-style set: temperature, seed, max_attempts_on_fail, env, use_native_tools, and max_tokens. Output is a single CONFIG.

Mechanically it builds an OllamaPromptDriver pointed at your localhost and wraps it in a DriversConfig. No API key anywhere - Ollama doesn't need one.

Why it's deprecated

The modern Ollama Drivers node covers the same ground and more: it adds an embedding driver (with a graceful dummy fallback if you leave it blank) and a bundled local vector store driver, so agent memory and RAG work without extra assembly. This old node is prompt-only - a memory-equipped agent required separate embedding/vector store nodes wired in by hand. It's kept around purely so old workflows keep loading. The old one also predates the fix that made the base_url/port settings actually flow through properly, so there were genuine bugs in early versions.

Installing it

Ships in the one pack:

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

or ComfyUI Manager → search "Griptape". Restart. For it to do anything you need Ollama running with a model pulled:

ollama run llama3

No keys to set - this is a zero-account node.

Should you use it?

Only to keep a legacy workflow alive. The switch to Ollama Drivers is a five-second swap on the same CONFIG socket, and you gain embeddings and memory for free. Building fresh, there's no reason to pick the deprecated node.

Common issues

• Connection refused - Ollama isn't running (ollama serve), or base_url/port don't match. Verify with curl http://127.0.0.1:11434.
• Model not found - prompt_model is free text and unvalidated; it must be a model you've actually pulled (ollama list to check).
• No memory - expected; the old node has no embedding driver. Add one manually or switch to Ollama Drivers.