Skip to content

Architecture

DPCS follows the canonical processing architecture:

Pipeline Contract
Canonical Object Model
Validation
Pipeline Plan
Capability Evaluation
Orchestrator Binding
Execution Runtime

Current implementation scope (through ROADMAP 0.13.0):

DPCS Document
Parser
Canonical Object Model
Validator (+ resolve nested DPCS refs, security, governance, versioning)
Diagnostics / DiagnosticReport
Pipeline Plan (deterministic, validation-gated, nested provenance)
Capability Evaluation (profile match, no plan mutation)
Orchestrator Binding (scaffolds + dpcs_semantics.json; capability-gated)

Also first-class (not mutating contracts):
  Compatibility analysis · Registry document validate · Registry HTTP client/server
  Conformance claims/profiles · Pipeline packages · JSON Schema / OpenAPI emit
  Language bindings (Python / WASM; see BINDINGS.md)
  Reports (Markdown/HTML/Mermaid/DOT) · Rich CLI formats · TUI inspector
  SPEC coverage matrix · Diagnostic catalog

Execution runtimes remain out of scope. Binding adapters emit implementation-defined scaffold artifacts for Airflow, Dagster, Prefect, Temporal, and Kubernetes, plus a structured dpcs_semantics.json encoding scheduling/QG/FS/execution/nesting. Registry networking uses the reference HTTP API (see ADR-0005). Language packages publish to PyPI (dpcs), npm (@eddiethedean/dpcs), and Wasmer (eddiethedean/dpcs).