Jump to content

Connect SuperML | Leeroopedia MCP: Equip your AI agents with best practices, code verification, and debugging knowledge. Powered by Leeroo — building Organizational Superintelligence. Contact us at founders@leeroo.com.

Implementation:Kubeflow Pipelines Pipeline Spec Protobuf

From Leeroopedia
Knowledge Sources
Domains Pipeline_Specification, API_Schema
Last Updated 2026-02-13 14:00 GMT

Overview

Auto-generated Go protobuf bindings for the complete KFP v2 pipeline specification schema, defining the canonical data model for pipelines, components, tasks, artifacts, parameters, and deployment configurations.

Description

The pipeline_spec.pb.go file is the single most critical API definition in KFP. Generated by `protoc-gen-go` (v1.36.6) from `pipeline_spec.proto`, it defines the Intermediate Representation (IR) that the Python SDK compiles pipelines into and that the backend interprets to orchestrate workflows. At 6,480 lines, it encompasses the entire pipeline specification language with 5 enums and 110 message types.

Key type hierarchy:

  • PipelineJob — Top-level job submission container
  • PipelineSpec — Component graph definition with runtime parameters
  • ComponentSpec — Individual component (DAG or executor)
  • DagSpec — Task dependency graph within a component
  • PipelineTaskSpec — Individual task with caching, triggers, retries, iterators
  • PipelineDeploymentConfig — Container, importer, resolver, and custom job specs
  • ExecutorInput/ExecutorOutput — Runtime I/O for task execution
  • PlatformSpec — Platform-specific deployment configuration

Enums: PrimitiveType (DOUBLE, INTEGER, STRING), ParameterType (NUMBER_DOUBLE, NUMBER_INTEGER, STRING, BOOLEAN, LIST, STRUCT, TASK_FINAL_STATUS), PipelineStateEnum, TaskConfigPassthroughType.

Usage

This protobuf package is imported by the KFP backend to deserialize pipeline specifications submitted by the SDK, and by the KFP SDK compiler to serialize pipeline definitions into the IR YAML format. Every pipeline submitted to KFP is serialized using these types.

Code Reference

Source Location

Signature

package pipelinespec

// PipelineJob is the top-level pipeline execution container.
type PipelineJob struct {
    Name           string
    PipelineSpec   *PipelineSpec
    RuntimeConfig  *PipelineJob_RuntimeConfig
    // ...
}

// PipelineSpec defines the component graph and deployment config.
type PipelineSpec struct {
    Components       map[string]*ComponentSpec
    DeploymentSpec   *PipelineDeploymentConfig
    Root             *ComponentSpec
    // ...
}

// ComponentSpec defines a single component as either DAG or executor.
type ComponentSpec struct {
    InputDefinitions  *ComponentInputsSpec
    OutputDefinitions *ComponentOutputsSpec
    // OneOf: Dag (*DagSpec) or ExecutorLabel (string)
}

Import

import "github.com/kubeflow/pipelines/api/v2alpha1/go/pipelinespec"

I/O Contract

Inputs

Name Type Required Description
Proto descriptor []byte Yes Raw protobuf descriptor bytes for pipeline_spec.proto

Outputs

Name Type Description
Go types structs 110 message types and 5 enums for pipeline specification
PipelineJob struct Top-level pipeline job submission type
PipelineSpec struct Component graph and deployment config type
ExecutorInput struct Runtime input passed to task executors

Usage Examples

Deserializing a Pipeline Spec

package main

import (
    "github.com/kubeflow/pipelines/api/v2alpha1/go/pipelinespec"
    "google.golang.org/protobuf/encoding/protojson"
)

func parsePipelineSpec(jsonData []byte) (*pipelinespec.PipelineSpec, error) {
    spec := &pipelinespec.PipelineSpec{}
    err := protojson.Unmarshal(jsonData, spec)
    if err != nil {
        return nil, err
    }
    return spec, nil
}

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.
Principle
Implementation
Heuristic
Environment