core / action / schema v1
Ask an AI
Send a prompt to an AI model and use its answer in the workflow
Prerequisites
No connection prerequisite is declared for this node.
Parameters
| Parameter | Type | Description |
|---|---|---|
AI providerprovider | select Required Default: | Pick "Other" for LM Studio, Ollama, Groq, OpenRouter, Together, anything speaking the OpenAI API. Options: OpenAI (ChatGPT) (openai), Anthropic (Claude) (anthropic), Other / self-hosted (OpenAI-compatible) (custom) |
Server URLbaseUrl | text Conditionally required | The API root, ending in /v1, not the /chat/completions path. A local server also needs its host and port allowed by the Noria operator.Field conditions{
"visible": {
"key": "provider",
"values": [
"custom"
]
},
"required": {
"key": "provider",
"values": [
"custom"
]
}
} |
API keyapiKey | text Conditionally required | Stored with the workflow and never shown again. Most local servers need no key. Leave it empty.Field conditions{
"required": {
"key": "provider",
"values": [
"openai",
"anthropic"
]
}
} |
Modelmodel | text Conditionally required | Exactly as the provider names it: "gpt-5.4-mini" or "claude-sonnet-5". On your own server, leave it empty to use the model already loaded.Field conditions{
"required": {
"key": "provider",
"values": [
"openai",
"anthropic"
]
}
} |
Promptprompt | textarea Required | What you are asking. Use {{variables}} to feed in anything from earlier steps. |
InstructionssystemPrompt | textarea Optional | Optional. Sets the tone and rules for every answer: language, length, what never to do. |
Response length limitmaxTokens | number Optional | Ceiling on the answer length (default 800). A Discord message tops out near 2000 characters, roughly 500 tokens. |
Creativitytemperature | number Optional | 0 gives the same answer every time, 1 varies (default 0.7). Low for classifying or extracting, higher for writing. |
Inputs and outputs
Inputs
execute- Execute (input)
Outputs
done- Done (output)text- Answer (output)model- Model used (output)error- Error (output)
Output data schema
[
{
"key": "text",
"type": "string",
"label": "Answer",
"description": "What the model replied, as plain text.",
"example": "Sure: here is a welcome message…"
},
{
"key": "model",
"type": "string",
"label": "Model used",
"description": "The model that answered. Useful when the server picks one for you.",
"example": "gpt-5.4-mini"
},
{
"key": "finishReason",
"type": "string",
"label": "Finish reason",
"description": "\"stop\" when the model finished on its own, \"length\" when Max tokens cut it off.",
"example": "stop",
"optional": true
}
]Examples and troubleshooting have not yet been reviewed for this contract.
Version and compatibility
This reference follows the current authoring contract. Existing workflows retain their saved node revisions.
Contract identifier
1d791fefb51dab9e0f01888c572268a84a5014ba23d5aa1e88a9cc762216ae46