01
A DAG with contracts between steps
Every edge in the graph is a schema and freshness contract. A step that would emit a breaking change halts there, with the downstream impact listed, instead of failing in a dashboard two hops later.
Why Genedata · EngineGENEDATA / 01
The GeneFlow engine.
Ingestion, transformation, and scheduling run as one graph on one runtime. Each edge carries a contract, each release is staged before it is live, and a backfill tells you what it will cost before you press run. That is why a pipeline built here never needs a second tool to test it, schedule it, or undo it.
Shared context · Lineage · Governance
Connected
Your sources
Data sources
Business context
Policy & access
Connected intelligenceGenedata
Governance
Business impactDecisions & action
Shared contextLineageGovernance
02
Staged releases, one-click rollback
A pipeline change runs in staging against production-shaped data and is promoted when its contracts pass. Roll back to the previous release from the run history; the checkpoint is still there.
03
Backfill with the cost shown first
Pick a window, and the engine reports rows, partitions, and compute before the job starts. Replays are idempotent against current logic, so a catch-up never needs a hand-written script.
The graph
Source to decision without leaving the runtime.
A representative pipeline. The contract check sits on the edge, not in a separate test suite, so a failing assertion halts publication before the warehouse, the dashboard, or the agent sees the change.
33 live · 84 cataloguedConnectors
One runtimeExecution graph
Staged then promotedRelease model
Cost shown before runBackfill
The next step
One graph is the whole argument.
The reason teams end up with five tools is that each one stops at its own boundary. GeneFlow keeps the graph continuous from the first connector to the last dashboard — which is also why lineage, monitoring, and policy on the other proof pages are possible at all.
FAQ
The engine, answered.
What engineers ask when they are deciding whether the graph can replace their scheduler.
What is a contract between steps, concretely?
A declared schema plus freshness, volume, and uniqueness assertions on the output of a step. The engine checks the contract before the next step reads the data. Additive changes — a new nullable column — propagate; breaking ones stop the run at that edge and open a review listing every downstream consumer.
How does a release get to production?
A change is committed as a new pipeline version and runs in staging against a sample of production-shaped data. When its contracts pass, you promote it; the previous version stays addressable in run history, so rollback is a selection rather than a redeploy.
What does the backfill cost estimate include?
Rows and partitions in the selected window, the steps that will recompute (only changed partitions, not the full table), and the compute that implies under your plan. You see it before confirming, and the hard cap on your plan is the ceiling regardless.
Can we keep our existing scheduler during the transition?
Yes. Most teams run GeneFlow in parallel on one pipeline first, compare outputs against the existing job, and retire the old schedule once the contracts have been green for a few cycles. Nothing has to be cut over on day one.
Next proof
The graph is only useful if you can trust what it carries.
The integrity proof shows how every figure downstream traces back to a committed record through these same contracts.