DeepsetSQLMetadataRetriever

Run SQL queries against metadata stored in deepset AI Platform.

Basic Information

• Type: deepset_cloud_custom_nodes.retrievers.deepset_sql_metadata_retriever.DeepsetSQLMetadataRetriever
• Components it can connect with:
  • Input that produces SQL strings at runtime.
  • PromptBuilder or components that consume the JSON response.
  • AnswerBuilder when you want to return the JSON response to clients.

Inputs

Parameter	Type	Default	Description
sql_query	string		SQL statement that targets the workspace metadata tables.
cursor	Optional[str]	None	Pagination cursor returned by the API.
fetch_size	Optional[int]	None	Number of rows to fetch per page. Overrides the init default.

Outputs

Parameter	Type	Default	Description
retrieved_information	Dict[str, Any]		JSON decoded response returned by the deepset metadata SQL endpoint.

Overview

Use DeepsetSQLMetadataRetriever to aggregate metadata values, join metadata tables, or to paginate through large result sets by forwarding the cursor returned in the response. The component strips Markdown style fences from the query so you can pass SQL prompts without extra parsing.

Usage Example

Initializing the Component

components:
  DeepsetSQLMetadataRetriever:
    type: deepset_cloud_custom_nodes.retrievers.deepset_sql_metadata_retriever.DeepsetSQLMetadataRetriever
    init_parameters:
      workspace: ${DEEPSET_WORKSPACE_ID}
      timeout: 120
      fetch_size: 200

Using the Component in a Pipeline

Coming soon

Bear with us while we're working on a pipeline template with DeepsetSQLMetadataRetriever.

Parameters

Init Parameters

Parameter	Type	Default	Description
workspace	str		Workspace ID whose metadata tables you query.
deepset_api_key	Secret	env var DEEPSET_API_KEY	API key used for the SQL endpoint.
timeout	float	80.0	HTTP timeout in seconds for the SQL request.
fetch_size	Optional[int]	None	Default number of rows fetched per page when run() does not override it.

Run Method Parameters

Parameter	Type	Default	Description
sql_query	string		SQL statement to execute. You can provide raw SQL or a code block wrapped in ```sql fences.
cursor	Optional[str]	None	Cursor token passed back to the endpoint when you resume pagination.
fetch_size	Optional[int]	None	Rows per page for this invocation. Fall back to the init value when omitted.