Skip to content

Catalog Context Search

The catalog context search agent is designed to answer questions about the Alation data catalog. This includes finding relevant objects, understanding relationships between objects, answering questions about the objects, and more.

The agent follows this workflow:

  1. Receives a natural language request from the user
  2. Analyzes the question to determine the best search strategy
  3. Searches for relevant custom fields, datasources, and filters to narrow down the search
  4. Defines a search signature to specify which objects and fields of the objects to retrieve
  5. Performs a preliminary search for relevant catalog objects using the search signature
  6. Retrieves relevant object details from the preliminary search results
  7. Addresses the user request using the retrieved object details

Required:

  • message (string): The natural language question or request to send to the agent

The agent produces a series of thinking, tool call, tool return, and text blocks as it works through the user request. The final message, assuming no errors, is a string which addresses the user’s request.

The agent is configured with 6 tools:

Analyzes the user’s question and determines a search strategy to address the request.

Key parameters:

  • message: The user’s natural language question

Retrieves all datasources (paginated) in the Alation catalog.

Retrieves all custom fields defined in the Alation catalog, including information on how to use them.

Lists instructions for generating a good search signature.

Searches for catalog objects using a search signature containing object types, fields, and search phrases.

Key parameters:

  • search_signature: The search signature defining object types, fields, and search phrases

Final search tool that retrieves detailed information about catalog objects by their IDs and a specified set of fields.

Key parameters:

  • search_signature: The search signature defining object types and ids, and fields to retrieve
  • The agent will answer the question provided as text, not just provide a list of search results.
  • The agent may use multiple search steps to refine results before answering the question.
  • If you know which object types are relevant to your request, including that in the request can help improve results.