Skip to content

Providers

One executable contract model for every lakehouse — powered by LakeLogic and proven live across Databricks, Snowflake, BigQuery, AWS, DuckLake, and MotherDuck (with Microsoft Fabric validated and deploy-ready).

This is the OLC equivalent of the Terraform Registry: one canonical contract, rendered across a matrix of backends. The RideFlow data mesh — the same contract model throughout — has been run on each platform below; only the backend-owned execution settings (engine, format, catalog, storage) change. Every provider page shows the identical contract, the framework invocation for that backend, and what it materialized.

What's RideFlow?

RideFlow is the running example used throughout OLC: a realistic, fictional ride-hailing company (like Uber) — its data mesh spans 6 domains (marketplace, payments, operations, marketing, reference, shared) across 11 source systems (Stripe, Twilio, Zendesk, Checkr, Google Ads, HubSpot, and more), ~60 governed tables through a bronze → silver → gold medallion. It exists as one set of OLC contracts; each provider page runs those exact contracts on a different lakehouse. Think of it as the "hello world" that's big enough to be real — nothing here is platform-specific to any one backend.

The contract never changes between these pages. Only the engine, the format, and where the data lands do.

The matrix

Platform Engine(s) Format Scope & status Reference repo
DuckDB / DuckLake DuckDB DuckLake ✅ Live — full 6-domain mesh, ~59 tables Coming soon
MotherDuck DuckDB DuckLake ✅ Live — marketplace domain, 18 tables + snapshots Coming soon
Databricks Spark, Polars, DuckDB Delta ✅ Live — full mesh on UC; external ADLS Delta via Polars/DuckDB GitHub ↗
Snowflake Snowflake SQL Native tables ✅ Live — full mesh on a trial, + native Project GitHub ↗
BigQuery (GCP) BigQuery, Spark BigQuery native, Iceberg ✅ Live — 11/11 systems; 18 Iceberg tables Coming soon
AWS (Glue) Spark Iceberg ✅ Live — reference + marketplace; one Silver job on Spark 3.5 Coming soon
Microsoft Fabric Spark Delta ◑ Validated & deploy-ready — assign a capacity GitHub ↗

How to read 'scope & status'

✅ Live = the contracts were executed on that platform and materialized real tables (scope stated per row). ◑ Static preview (Fabric) = validated and deploy-ready; Fabric is capacity-based, so a live run expects an F-SKU (or trial) capacity assigned to the workspace — a normal customer prerequisite. Each provider page states exactly what ran — no silent gaps.

Each reference repo is the reproducibility artifact for that backend — the exact contracts, the framework invocation, and the run output. They are made public alongside the OLC release.

What "same contract" means here

Each provider page starts from a shared RideFlow silver contract like this:

version: 1.0.0
info: { title: Trips, table_name: silver_rideflow_trips, target_layer: silver }
model:
  fields:
    - { name: trip_id,   type: string, required: true }
    - { name: rider_id,  type: string, required: true }
    - { name: fare_amount, type: float }
    - { name: rider_email, type: string, pii: true, masking: partial }
primary_key: [trip_id]
quality:
  row_rules:
    - { name: fare_non_negative, sql: "fare_amount >= 0" }
materialization: { strategy: merge }     # format chosen per-platform

The only per-platform choices are the ones a backend legitimately owns:

flowchart LR
    K[One RideFlow contract set] --> D["DuckDB · DuckLake · local/S3/GCS/ADLS<br/>(DuckDB / DuckLake)"]
    K --> MD["DuckDB · DuckLake · MotherDuck<br/>(MotherDuck)"]
    K --> DB["Spark · Delta · Unity Catalog<br/>(Databricks)"]
    K --> SF["Snowflake · native · Task DAG<br/>(Snowflake)"]
    K --> BQ["BigQuery · native + Iceberg · GCS<br/>(BigQuery / GCP)"]
    K --> FB["Spark · Delta · OneLake<br/>(Microsoft Fabric)"]
    K --> AW["Spark · Iceberg · Glue + S3<br/>(AWS Glue)"]

Special configurations

Like Terraform's per-provider "special configuration" notes, each backend has a small number of real quirks the framework handles for you. The short version:

Platform Notable per-backend handling
Snowflake UPPERCASE identifier casing + dedup ordering; TO_VARCHAR casts; native Project from repo root
BigQuery Backtick-quoted identifiers; client location; TEMP needs a session; CONCAT_WS / TO_DATE rewrites
DuckLake ATTACH 'ducklake:…' (DATA_PATH …) locally; CREATE DATABASE … (TYPE DUCKLAKE) on MotherDuck
AWS Glue Glue Catalog-registered Iceberg; checkpoint-before-merge to avoid Catalyst plan explosion
Azure/Fabric OneLake schema-enabled Lakehouse; REST/OneLake deploy via az token

Each is covered on the platform's page.