API Overview¶

This page is a lookup map for the public SDK API. The Core API guides explain when and how to use each record type; this index points to the relevant declarations and runtime helpers.

Need	Primary API	Guide
Persist one external or source value	`@json_artifact`, `@csv_artifact`, `@parquet_artifact`, and other Artifact decorators	Artifacts
Declare reusable work and publish its real invocation	`@computation`, `computation_record`, `OclpRun.outputs_for()`	Computations and Executions
Record a quality or acceptance conclusion	`@evidence`, `evaluate_evidence`	Evidence
Package exact Artifacts	`ComputationArtifactSet`, `@artifact_set`, `assemble_artifact_set`, `artifact_set_input`	Artifact sets
Observe related real Executions	`@run`, `observe_run`, `OclpRun`, `active_run`	Runs
Bind code and persist records	`GitSource`, `ArtifactSource`, `ServiceSource`, `OpaqueSource`, `LocalArtifactPublisher`	Sources and publishing
Parse, digest, and validate portable records	`parse_record`, `canonical_json_bytes`, `record_digest`, validation helpers	Canonical records and validation

Core record models¶

The immutable Pydantic models live in oclp.models:

from oclp.models import (
    Artifact,
    ArtifactSet,
    Computation,
    Evidence,
    Event,
    Execution,
    RecordReference,
)

Their field semantics and conformance requirements are defined by the OCLP specification. Use the SDK declarations for ordinary application code; instantiate model objects directly when implementing a store, ingesting records, or performing a low-level integration.

Optional integrations¶

Dagster activates canonical OCLP declarations inside native Dagster assets.
MLflow mirrors OCLP observations for experiment tracking.
DuckDB catalog provides local record and Artifact-location storage.