Griptape Agent Config: Google [DEPRECATED]

You'll usually find this node in an old downloaded workflow rather than by searching for it fresh, because it's the deprecated predecessor of "Griptape Agent Config: Google Drivers." The [DEPRECATED] in the name is the pack telling you it's still there so old workflows keep loading, but it's not where you should build new stuff. It works - it just does less.

What it is

This is the original "Google Structure Config": a single config node that points a Griptape agent's prompt driver at a Gemini model. One required input - prompt_model, which is a free-text model name - plus the shared temperature, seed, max_attempts_on_fail, use_native_tools, max_tokens, an env input for environment variables, and api_key_env_var (default GOOGLE_API_KEY). It outputs a single CONFIG that feeds an Agent.

Why it's deprecated

The modern replacement, Google Drivers, does the same prompt-driver job and adds an embedding driver with a bundled local vector store - which is what gives agents actual memory and RAG. This old node only wires up the prompt driver, so any workflow that wants memory had to bolt on embedding/vector store nodes separately. The pack has been migrating everything to the "Drivers" config family; this is the tail end of that process.

The practical differences to care about:

• This node takes prompt_model as a raw string you must type (e.g. gemini-1.5-pro); the new one gives you a dropdown with current models like gemini-2.0-flash.
• No embedding driver, no vector store - no out-of-the-box agent memory.
• It uses the older single-config plumbing instead of the per-provider DriversConfig.

Installing it

You don't install "this node" - it ships in the one pack:

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

or ComfyUI Manager → search "Griptape" → install. Restart, set your GOOGLE_API_KEY (Settings → Griptape or env var), and old workflows that reference this node will load and run.

Should you use it?

Only if you're maintaining a workflow that already uses it. The effort of switching is tiny - the new Google Drivers node occupies the same CONFIG socket, and you re-enter the model name once from a dropdown. If you're starting fresh, don't add this node deliberately; grab the Drivers version and get memory for free. The only scenario where the old node is arguably still convenient is a workflow where you specifically don't want an embedding driver at all and prefer the slimmer config object.

Common issues

• Missing model name - this node doesn't validate the string; a typo in prompt_model gives you an API error instead of a dropdown to pick from.
• Agent has no memory - not a bug, just the old design. Add embedding/vector store nodes manually or switch to the Drivers config.
• Key not found - same [WARNING]: GOOGLE_API_KEY is not set console message as everywhere in the pack; check Settings → Griptape and the env var name spelling.