Selecting objects for AI code conversion

The Select objects page offers a comprehensive view of the database code conversion workload and the corresponding status of each code unit.

  1. You can perform the following actions on the database objects shown in a tree view under the Name column:

    • Select the database objects for AI code conversion.

    • Mark objects as VERIFIED BY USER to exclude them from subsequent runs of the AI code conversion process. Only these objects will be considered as valid for deployment.

    • Filter the database objects by type, database, schema and status.

  2. The Selection Summary indicates the number of selected objects, the associated dependencies, and the estimated time and cost for the AI code conversion. Note that the total number of objects includes the object selected and its dependencies. The estimated time and costs are depend upon the size of the selected objects and token usage based on historical benchmark data.

  3. The Breakdown section shows all objects grouped by their conversion status as below:

    • Converted successfully - The object was successfully converted with deterministic conversion (non AI code conversion) and is ready to be deployed to Snowflake.

    • Has issues - The object has issues from either deterministic or AI code conversion and cannot be deployed without fixes.

    • Suggested fixes - The AI code conversion process has generated suggested fixes and requires user review.

    • Verified by AI/AI converted - The converted code was found to produce equivalent results to the source and is ready for user review.

  4. The process of running AI code conversion is meant to be iterative and can be run multiple times on the same object until the expected accuracy is achieved. Select Go to latest conversion at the bottom of the page to view the results of the last AI code conversion.

Language: English