Actions51
- Genie Actions
- Databricks SQL Actions
- Unity Catalog Actions
- Model Serving Actions
- Files Actions
- Vector Search Actions
- Jobs Actions
- Clusters Actions
Model Serving → Query Endpoint
AI-generatedSummary
Send a JSON-formatted query to a specified Databricks model serving endpoint, automatically detecting the correct invocation URL and input format from the endpoint's OpenAPI schema.
Inputs
- Endpoint (required) — The model serving endpoint to query. Can be selected from a list, entered by name, or by URL. The node fetches the endpoint's OpenAPI schema to determine the correct invocation URL and input format.
- Request Body (required) — The JSON-formatted request payload to send to the serving endpoint. The input format is validated at runtime against the endpoint's OpenAPI schema. Example formats depend on the model (e.g., chat messages).
Output shape
a single JSON object containing the endpoint's response merged with metadata about the invoked endpoint, including detected input format and invocation URL, paired with the corresponding input item.
The node performs an HTTP POST to the model endpoint's invocation URL, which is dynamically retrieved from the endpoint's OpenAPI schema. If the OpenAPI schema is unavailable, the node falls back to a default invocation URL pattern. Possible input formats are automatically detected by inspecting the endpoint schema. Errors are raised if underlying HTTP requests fail.
Examples
Example 1: Query a deployed machine learning model endpoint with a chat message.
Configure the 'Endpoint' parameter by selecting from the list of available serving endpoints or specifying the endpoint name or URL. Provide the request body in JSON format matching the model's expected input schema, such as a chat message with role and content fields.