Consort: Open-Source Framework for Test-Driven Development on Branching Databases
Databricks Field Engineering has launched Consort, an open-source agentic development framework designed to enable test-driven development directly on real, isolated branching databases like Lakebase Postgres. This framework addresses the historical challenge of integrating databases into modern development workflows, allowing for inner-loop integration testing without mocks, unified code and schema deployment, and earlier detection of production issues. It models development team roles as "agents" within a structured workflow, guided by a "conductor" to ensure tests pass against live data. Consort is community-supported and includes a VS Code plugin for visualizing development and an observability dashboard.
- →Agentic development workflow with defined roles
- →Visualizing development and supporting parallel experimentation
- →Consort: An open-source agentic framework for TDD on branching databases
- →Improving integration testing and production stability
Features (2) ›
- Agentic development workflow with defined roles
Consort models development team roles (e.g., product owner, architect reviewer, DBA, test strategist) as 'agents' led by a 'conductor' (code-based state machine). This workflow ensures agents interact with real database branches, preventing goalpost shifting and maintaining focus through scoped context packages and dedicated artifact directories.
- Visualizing development and supporting parallel experimentation
The framework includes a VS Code plugin to display paired Git and Lakebase branches, showing code and schema changes in a single diff view. An observability dashboard provides live insights into each agent's actions and produced artifacts, also supporting parallel experiments on isolated database branches for design exploration.
Enhancements (1) ›
- Improving integration testing and production stability
The framework replaces mock objects with isolated, live database branches for inner-loop integration testing, regardless of data size. It enables full test suites to run against fresh database branches on every pull request and merge, catching regressions and conflicts pre-deployment and simplifying schema change promotion.
Notes (1) ›
- Consort: An open-source agentic framework for TDD on branching databases
Consort is an open-source agentic development framework from Databricks Field Engineering designed for test-driven development directly on branching databases. It aims to integrate databases into modern software practices by treating them like code repositories, addressing the challenge of rigid database monoliths.
https://www.databricks.com/blog/introducing-consort-test-driven-development-branching-database
Related releases
- Databricks details Adaptive Instructed-Retriever for efficient AI search Databricks Blog ·
- Databricks Go SDK v0.178.0 Enhances ML Features and Adds AI/Sandbox APIs Databricks Go SDK Releases ·
- Databricks SDK for Java v0.154.0 Enhances Feature Engineering & ML Policy APIs Databricks Java SDK Releases ·
- Databricks Python SDK v0.137.0 Enhances Feature Engineering and ML Metadata APIs Databricks Python SDK Releases ·
- Zepto Scales AI Customer Support with Evaluation-First Agents on Databricks & MLflow Databricks Blog ·
- Databricks Lakebase Expands PCI-DSS and HITRUST Compliance to All AWS Regions Databricks Release Notes ·