Build on Lakebase,
not on plumbing.
Examples and guides to accelerate your Databricks Lakebase projects. Postgres you already know, integrated with your lakehouse
LTAP, branching, autoscaling, scale-to-zero
Patterns you can deploy today.
Agents
02 examplesKnowledge-graph RAG on Lakebase
pgvector seed, recursive-CTE graph traversal
Augment RAG with a knowledge graph stored in Lakebase. A pgvector semantic seed finds entry nodes, a recursive CTE traverses relationships out to k hops, and a blended score ranks the expanded context — surfacing connections flat retrieval misses.
Durable memory for agents
Chainlit history + pgvector semantic recall
Give an agent persistent memory on one Lakebase store: short-term conversation history via Chainlit’s data layer, long-term facts, and semantic recall over them with pgvector — the app owns its schema, so access survives redeploys.
Developer Experience
04 examplesShip schema changes with confidence
branch-per-PR, AI impact review, exactly-once deploy
A GitHub CI/CD workflow that forks a Lakebase branch off production for every PR touching migrations, runs the changed SQL there, and posts an AI impact review before merge — then applies migrations to production exactly once.
Consort — enforced spec-first TDD on Lakebase branches
role agents, deterministic gates, honest-green
An agentic development framework: a Scrum team rendered as role agents runs spec-first, test-driven development where every "green" is a real test on a live Lakebase branch, driven by a deterministic state machine with human-approval gates and immutable tests.
Paired Git + Lakebase branching in your IDE
VS Code / Cursor, branch pairing, credential rotation
A VS Code / Cursor extension that replaces built-in Git source control with a unified Git + Lakebase provider, pairing every code branch with a Lakebase database branch so code and schema travel together through review, CI/CD, and merge.
The portable Lakebase branching + SCM engine
library API + CLIs, paired-branch state machine
The portable engine behind the SCM extension and Consort: database branching, the paired-branch SCM workflow state machine, credential minting, and schema migration — consumable as a TypeScript library and as CLIs.
Apps
02 examplesFastAPI backend on Lakebase
Databricks App, OAuth rotation, scale-to-zero pooling
A Databricks App that serves a Lakebase synced table through a FastAPI REST API — with automatic OAuth token rotation, a scale-to-zero-aware connection pool, and fully bundle-driven provisioning of the project, endpoint, catalog, and synced table.
Semantic caching for Genie Spaces
pgvector cache, multi-space routing, feedback loop
A Databricks App that puts a pgvector semantic cache in front of Genie Spaces, so a reworded question re-executes the SQL that already worked instead of regenerating it — with per-user memory, supervisor routing across Spaces, and a feedback loop that promotes thumbs-up answers into the cache.
Data
02 examplesOnline feature serving
synced tables → model serving
Serve features to models with low latency by syncing lakehouse tables into Lakebase — no bespoke online store to build or operate.
Lakehouse → operational store
managed sync, no custom pipelines
Push analytical results back into Lakebase for low-latency serving to applications, without building and maintaining reverse-ETL pipelines.
Operational and analytical, one source of truth
Stop replicating data between an operational database and your lakehouse. Lakebase brings a fully-managed Postgres engine directly into the platform — so the same governed data powers your apps, agents, and analytics.