Migrating from a legacy knowledge warehouse is a fancy enterprise, requiring groups to investigate decades-old code, translate proprietary dialects, migrate large datasets, and reconcile variations between methods. Databricks is making it dramatically simpler emigrate workloads from legacy knowledge warehouses to the Lakehouse with Genie Code, Databricks’ AI coding agent.
We’re excited to announce the brand new agentic converter in Genie Code. The agentic code converter interprets proprietary dialects to open ANSI SQL, beginning with T-SQL, Snowflake, Redshift, Oracle, BigQuery, and Teradata. It transforms an information warehouse migration from a venture your crew staffs and manages to 1 you configure, launch, and monitor.
Automated migration planning
To see how the agentic converter works in follow, we’ll stroll by way of a proof of idea changing a set of T-SQL saved procedures to ANSI SQL.
To start out the migration, we create a migration venture, a brand new function in a Databricks workspace. Migration tasks give groups a central hub to handle supply information, monitor conversion progress, and collaborate all through the migration. We identify our venture Migration Mission – POC, set the supply dialect to T-SQL, the goal to ANSI SQL, and select a vacation spot folder for the transformed information. As soon as created, we will populate the venture with supply SQL information, which had been beforehand uploaded to the workspace. Every file seems within the venture with its file kind, strains of code, and migration standing:
Genie Code analyzes and scores every file’s complexity, surfacing an evaluation in the best panel. In our instance, mixed_5cats_sp_string_agg.sql scores low complexity as a result of it accommodates solely SQL options that map cleanly to ANSI SQL. For a full scale migration, groups can use the complexity rating to prioritize and deal with easier information first.
Genie Code additionally generates lineage, mapping each relationship between objects within the legacy property — tables, views, and saved procedures. The lineage graph reveals that sps_sp_update_from.sql and sps_sp_pivot.sql don’t have any shared dependencies, that means they are often safely migrated independently.
In a full-scale migration, complexity and lineage give groups a transparent image of what emigrate, in what order, and what wants to maneuver collectively.
Totally agentic code conversion
After analyzing the supply code, we will begin to convert code. Once we click on Run, Genie Code analyzes every script’s T-SQL and launches swarms of subagents to transform information in parallel. Every subagent iteratively fixes errors, validating each syntax and semantic intent to make sure the transformed code retains the unique enterprise logic and efficiently parses.
As soon as full, the transformed information are written to the vacation spot folder and color-coded by standing. In our proof of idea, 6 of 8 information had been efficiently transformed. Nevertheless, mixed_5cats_sp_string_agg.sql and mixed_6cats_sp_string_agg.sql want additional overview. Viewing the best panel, we will see the checklist of to-do’s required to repair the script. Genie Code explains that the flagged saved procedures should be certified by a 3 half identify in Unity Catalog (catalog.schema.sp_string_agg).
By clicking on the file, we will open the native SQL editor to discover the side-by-side distinction. We are able to make the repair manually within the editor or codify a conversion rule by making a customized talent in Genie Code. Once we’re able to run the complete migration, Genie Code will apply the rule, on this case giving saved procedures a 3 half identify in Unity Catalog, mechanically throughout the complete codebase. Customized expertise are additionally helpful for broader customization, like conforming to an permitted ETL sample or following inside naming conventions.
For groups seeking to lift-and-shift, Databricks ships the complete suite of enterprise SQL options. Multi-Assertion Transactions, Temp Tables, and Saved Procedures — options generally utilized in legacy knowledge warehouses – are all obtainable in Databricks, so there is not any must re-engineer your logic to suit a brand new platform.
Get began with the Agentic Converter right now
The agentic code converter is the subsequent era of information warehouse migration tooling, constructing on Lakebridge’s success at scale. Since its launch at Knowledge and AI Summit 2025, Lakebridge’s conversion tooling has helped over a thousand prospects undertake the Databricks lakehouse:
“Lakebridge made our migration to Databricks extra predictable and environment friendly. It dealt with the conversion of our legacy saved procedures into Spark Declarative Pipelines, which reduce down a number of the handbook work and helped us transfer sooner towards a consolidated open Lakehouse.”—Shane Irons, Sr. Director of Info Administration
As a subsequent step, we’re increasing the agentic converter to assist widespread legacy ETL sources and new goal dialects. We additionally plan to carry knowledge migration and reconciliation into the migration venture expertise, with a local Lakeflow Join integration emigrate knowledge and automatic instruments to validate that migrated knowledge matches the supply earlier than cutover.
Able to migrate? Join along with your Databricks account crew to evaluate your legacy property and check out the agentic code converter right here.