Griptape Agent Config: Azure OpenAI [DEPRECATED]

The deprecated Azure OpenAI config is the old pre-driver way to point a Griptape agent at models served through Azure's OpenAI service. It's marked [DEPRECATED] for the usual reason - the pack replaced it with the driver-based Azure OpenAI Drivers config - but if you're maintaining an existing workflow (or your org is locked into Azure OpenAI and the old node is what you have), it still runs.

Azure OpenAI is the corporate-flavored version of OpenAI: you don't call gpt-4o by name, you call the deployment your admin created for it. That's the single most important thing to understand about this node and its successor. The dropdown's model names (gpt-4o, gpt-4, gpt-3.5-turbo-16k, gpt-3.5-turbo) are just labels for the driver; what actually gets sent to Azure is the prompt_model_deployment_name (default gpt4o) plus your endpoint. If the deployment name is wrong, every call fails regardless of which model you picked.

How it works

The node builds an AzureOpenAiChatPromptDriver from the model label and deployment name, reads AZURE_OPENAI_API_KEY and AZURE_OPENAI_ENDPOINT from the env vars you name, and wraps the driver in an AzureOpenAiDriversConfig. Output is a CONFIG for agent creation. Nothing exotic - it's the OpenAI driver pattern with Azure's endpoint/deployment indirection bolted on.

Inputs and outputs

• prompt_model - required dropdown: gpt-4o (default), gpt-4, gpt-3.5-turbo-16k, gpt-3.5-turbo.
• prompt_model_deployment_name - required; the actual deployment name in your Azure OpenAI resource. Default gpt4o.
• temperature, seed, max_attempts_on_fail, env, use_native_tools, max_tokens - the standard settings.
• api_key_env_var - env var name, default AZURE_OPENAI_API_KEY.
• azure_endpoint_env_var - env var name for the endpoint, default AZURE_OPENAI_ENDPOINT.
• CONFIG - the output.

Note the description in ComfyUI references AZURE_OPENAI_ENDPOINT_3 and AZURE_OPENAI_API_KEY_3 - that's leftover from an older setup; the actual defaults in the node are the non-_3 names.

Install and credentials

Same pack - ComfyUI Manager (search "Griptape") or:

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

Endpoint and key come from your Azure OpenAI resource in the Azure portal (the README links Microsoft's endpoint-switching guide). Set both in Settings → Griptape or your environment. Standard pack caveats: griptape[all] and its torch dependency (cu121 reinstall on Nvidia if ComfyUI breaks).

Common gotchas

Deployment-name mismatch is the #1 Azure failure - you'll get a 404-ish error even with a valid key because the deployment doesn't exist under that name. Confirm the exact name in the Azure portal's "Deployments" list. Also, gpt-3.5-turbo-16k and friends are old - your Azure admin may have decommissioned them, so a dropdown choice that's "valid" can still fail at runtime. And as with every deprecated config here: build new workflows on the Azure OpenAI Drivers node instead, which adds image generation and embedding drivers and handles the deployment names for each.