Natural Language Data Source Matching (Preview)

oracle_analytics-find_matching_datasources

This tool requires datasets to be indexed before they can be matched to natural language questions. Datasets are indexed using the same indexing process that Oracle Analytics Cloud uses for its AI features.

Input Schema

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "type": "object",
  "properties": {
    "nl_question": {
      "type": "string"
    }
  }
}

Finds candidate subject areas or datasets for a natural-language analytics question.

Use this when the user starts with a business question and does not know which data source to use. If it can't find a good match, fall back to oracle_analytics-search_catalog.

Import Parameters

Parameter	Type	Notes
`nl_question`	string	Natural-language analytics question.

Example

{
  "jsonrpc": "2.0",
  "id": 20,
  "method": "tools/call",
  "params": {
    "name": "oracle_analytics-find_matching_datasources",
    "arguments": {
      "nl_question": "Which products had the highest revenue last quarter by channel?"
    }
  }
}

Typical Response Fields

Field	Notes
`name`	Data source name.
`displayName`	User-facing name.
`description`	Business description where available.
`score`	Intent-match score.

Recommended Flow

Call oracle_analytics-find_matching_datasources.
Re-rank candidates using descriptions and business context.
Call oracle_analytics-describe_data for the best candidate.
Build and validate Logical SQL.
Execute with oracle_analytics-execute_logical_sql.