Migrating from a legacy knowledge warehouse is a posh enterprise, requiring groups to research decades-old code, translate proprietary dialects, migrate huge 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 a knowledge warehouse migration from a undertaking your workforce staffs and manages to 1 you configure, launch, and monitor.
Automated migration planning
To see how the agentic converter works in observe, 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 undertaking, a brand new characteristic in a Databricks workspace. Migration initiatives give groups a central hub to handle supply recordsdata, monitor conversion progress, and collaborate all through the migration. We title our undertaking Migration Challenge – POC, set the supply dialect to T-SQL, the goal to ANSI SQL, and select a vacation spot folder for the transformed recordsdata. As soon as created, we will populate the undertaking with supply SQL recordsdata, which have been beforehand uploaded to the workspace. Every file seems within the undertaking with its file sort, strains of code, and migration standing:
Genie Code analyzes and scores every file’s complexity, surfacing an evaluation in the precise panel. In our instance, mixed_5cats_sp_string_agg.sql scores low complexity as a result of it comprises 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 less complicated recordsdata first.
Genie Code additionally generates lineage, mapping each relationship between objects within the legacy property — tables, views, and saved procedures. The lineage graph exhibits that sps_sp_update_from.sql and sps_sp_pivot.sql haven’t 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 recordsdata 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 recordsdata are written to the vacation spot folder and color-coded by standing. In our proof of idea, 6 of 8 recordsdata have been efficiently transformed. Nonetheless, mixed_5cats_sp_string_agg.sql and mixed_6cats_sp_string_agg.sql want additional evaluate. Viewing the precise panel, we will see the record of to-do’s required to repair the script. Genie Code explains that the flagged saved procedures should be certified by a 3 half title 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 ability 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 title in Unity Catalog, mechanically throughout your complete codebase. Customized abilities are additionally helpful for broader customization, like conforming to an authorized 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 out there 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 technology 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 minimize down a variety of the guide work and helped us transfer sooner towards a consolidated open Lakehouse.”—Shane Irons, Sr. Director of Data 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 convey knowledge migration and reconciliation into the migration undertaking 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 workforce to evaluate your legacy property and take a look at the agentic code converter right here.
