Processor

The processor is the part of a data product that executes the stream logic you model in Spine: operators on streams, joins, keyed state, and publishing to outlets (and, where defined, support for key–value and other product surfaces). It sits between ingests and outlets / APIs.

In source, the processor is the compiled realization of the flow (and related declarations) in your .spine files, driven by the Data Product Manifest and libraries you attach at build time. At runtime, it applies correctness and ordering guarantees the platform enforces: events are processed in a well-defined way so downstream consumers and operational tooling see a coherent, replayable history.

Product-wide order — The processor determines a single global order for the entire data product (how work from ingests and related inputs is interleaved and sequenced; see also Spine: data product pipelines). For any given message, its position in that order is set in processing and does not change after that, including in replay and other re-execution paths.

The processor is not a separate service you call directly; you interact with the product through ingests (writes), outlets (stream reads), and unary APIs (request/response) as those are published for the product.

See also — Ingests · Outlets · Data product lifecycle · Spine: data product pipelines