Move · Pipeline orchestrator
Revoluzen Flow
An open-source orchestrator you install with one command. Draw a pipeline on a canvas or write it in SQL and Python, schedule it, and get a run history detailed enough to answer questions about it later.
extract_orders
1,303,343 rows
extract_refunds
retry 1 · 41,207 rows
conform
1,261,118 rows
assert_contract
37 checks · 0 findings
load_gold
writing · 38%
Every step on the canvas is a row in the run — inputs, outputs, retries and the receipt it ends in.
- License
- Apache-2.0
- Version
- v1.0.1
- Install
- pip install revoluzen-flow
- Binds
- 127.0.0.1:8001 (loopback)
Who it is for
For teams who want their pipelines on their own machines — and who are tired of paying per row for a tool that copies production data somewhere else.
The move plane
Each product is adoptable on its own; together they share one catalog, one permission model and one run history.
01
Move
- Flow
- Flow Enterprise
02
Govern
- Catalog
03
Answer
- Studio
04
Operate
- Console
- Mobile
05
Prove
- Evidence
Draw it, or write it
The canvas and the code are the same pipeline, not two products. An analyst can lay out a flow visually; an engineer can drop into a SQL or Python step inside it without leaving the graph.
- 0140+ step types on the canvas, from extract and load to transform, branch and assert
- 02SQL steps run against your own warehouse; Python steps run in your own environment
- 03Every step shows its input and output so a failure is readable, not archaeological
- 0427 starter templates covering the common shapes
Connect what you actually have
Databases, object stores, files and SaaS REST APIs. The open-source build covers the open ecosystem end to end; the enterprise systems of record are the commercial boundary.
- 01100+ connector types in the open-source build
- 02Postgres, MySQL, SQL Server, SQLite, DuckDB, MongoDB and the usual object stores
- 03SaaS sources defined as declarative REST manifests, so a new one is config, not a fork
- 04Connection secrets encrypted at rest — no external key service required to start
Re-runs that do not ruin your warehouse
Most orchestration pain is not the happy path, it is the second attempt. Resume and backfill semantics are built in rather than left as an exercise for the operator.
- 01Per-step resume: a failed run continues from the failed step, not from the beginning
- 02Idempotent backfills, so re-processing a window does not duplicate rows
- 03Retry policies per step, with the retry recorded in the run history
- 04Scheduled and manual runs share one execution path, so they fail the same way
Steward: the checks you would have written eventually
A reliability layer that inspects your pipelines and runs, and raises findings across architecture, pipeline, node, connector, data, governance and cost.
- 0137 finding kinds across 7 levels
- 02Flags the classics: retry storms, partial outputs, cast failures, credential sprawl
- 03Findings link back to the step that caused them
- 04A rules engine for the checks that are specific to your business
The details, in one table
The facts a technical evaluator checks first, without having to read a datasheet.
| License | Apache-2.0 |
| Package on PyPI | revoluzen-flow |
| Install | pip install revoluzen-flow |
| Default storage | SQLite (file-based, no server) |
| Default bind | 127.0.0.1:8001, loopback only |
| LAN access | Opt-in via REVOLUZEN_FLOW_ALLOW_LAN |
| Secrets at rest | Fernet symmetric encryption, always on |
| Telemetry | None |
| Runtime | Python backend, browser UI |
Run it now
pip install revoluzen-flow && revoluzen openNothing to sign up for. The server binds to loopback, stores state in a local file, and sends us nothing.
Open source, and not
Published rather than discovered. If a capability is commercial, it says so here and on the page where you would look for it.
In the open-source build
- The full canvas, scheduler, run history and Steward
- 100+ connector types across the open ecosystem
- The Copilot, pointed at a model you choose
- Single-node deployment with SQLite storage
Licensed separately
- Change data capture via Debezium
- Enterprise and SaaS systems-of-record connectors
- Postgres metastore, workspaces and environment promotion
- Approval gates, external execution providers, HA topologies
What this does not do yet
You would find these out eventually. Finding them out now is cheaper for both of us.
Connector maturity is published per connector — most are SDK-validated, some are integration-verified, and none are yet marked production-ready. Check the one you need before you plan around it.
Change data capture is not in the open-source build; it is part of Flow Enterprise.
The enterprise systems of record (SAP, Oracle, Teradata, DB2, Salesforce, Workday, NetSuite, ServiceNow, Dynamics 365) are commercial connectors.
SQLite storage is the right default for one machine. A multi-user production deployment should move the metastore to Postgres, which is a Flow Enterprise capability.
Related products
Enterprise edition
Revoluzen Flow Enterprise
Everything in Flow, plus the parts an enterprise platform team asks for on day one: CDC, enterprise sources, workspaces and release control.
Read moreGoverned lakehouse
Revoluzen Catalog
A sovereign lakehouse and catalog: medallion layers, dimensional models, lineage and a governed SQL endpoint over open table formats.
Read moreAnalytics & reporting
Revoluzen Studio
Dashboards, pixel-accurate reports and a governed query layer, running inside your network against your own warehouse.
Read moreRun it yourself, then decide.
Revoluzen Flow is Apache-2.0 and installs in one command. If it fits, we can talk about the parts that are not open source.