The cds-data suite

cds-data is a set of three composable SAP CAP plugins. You usually reach for one of them by intent; the other two are either composed automatically or used standalone.

Plugin	You reach for it when…	Surface
cds-data-federation	You have a remote service entity you want available in your app — live or synced.	@federation.delegate / @federation.replicate on consumption views.
cds-data-pipeline	You need a scheduled, traceable READ → MAP → WRITE job — from any source to any target, in code.	cds.connect.to('data-pipeline').addPipeline({ … }) + management API + Pipeline Console.
cds-data-materialization Experimental	You want a local aggregate snapshot from a group by projection.	@materialize.snapshot. Not yet released.

How they compose

The pipeline is the engine. The two annotation plugins sit on top of it and generate pipeline registrations for you — you never call addPipeline yourself when you use annotations.

@federation.delegate ─── live proxy ──────────────▶ remote service   (no pipeline)
@federation.replicate ─┐
                       ├─▶ cds-data-pipeline ──▶ your database
@materialize.snapshot ─┘        (engine)

• @federation.replicate turns each annotated consumption view into one entity-shape pipeline (row-preserving UPSERT into a local table).
• @materialize.snapshot turns each group by projection into one query-shape pipeline (snapshot rebuild of an aggregate).
• cds-data-pipeline runs standalone too — register pipelines programmatically for any source/target the annotation layers don't cover.

Both annotation plugins declare cds-data-pipeline as a peer dependency and fail loudly at startup if it is missing.

See Architecture for the layering in detail, and Terminology for the exact meaning of federation, delegation, replication, materialization, and pipeline.

Feature highlights per plugin

cds-data-federation

• Delegate — transparent live proxy; reads forwarded to the remote, optional CUD opt-in.
• Replicate — scheduled sync into the local DB for offline access, analytics, and full SQL.
• Cross-service $expand and navigation — stitch local and remote entities across service boundaries.
• Caching — optional response cache (cds-caching) or entity cache (SQLite snapshot via the pipeline).
• Multi-tenancy — per-tenant replication and entity cache under @sap/cds-mtxs.

→ Federation features · Annotations

cds-data-pipeline

• Sources — OData V2/V4, HCQL, REST (with pagination + delta), CQN, and custom adapters.
• Targets — local DB (UPSERT/INSERT/truncate), remote OData, and custom adapters.
• Delta tracking — timestamp, key, and datetime-fields strategies with a persisted high-watermark.
• Scheduling — in-process (spawn), cross-instance single-winner (queued), or external trigger.
• Observability — Pipelines / PipelineRuns tracker, management OData API, and the Pipeline Console monitor UI.
• Event hooks — before / on / after for every phase (START → READ → MAP → WRITE → DONE).

→ Pipeline features · API reference

cds-data-materialization Experimental

• @materialize.snapshot on group by projections — persist scheduled rollups locally.
• Composes the pipeline for the query-shape run; stage remote data with @federation.replicate first, then aggregate.

→ Materialization features

Common ground

• Terminology — the shared vocabulary across all three plugins.
• Architecture — the engine-vs-annotation layering and the consumption-view contract.
• CDS 10, HCQL and MCP — how the June 2026 CAP release composes with the suite.