Implementation:Kubeflow Pipelines MLMD Store Protobuf

Knowledge Sources	Kubeflow_Pipelines ML Metadata
Domains	Metadata_Management, API_Schema
Last Updated	2026-02-13 14:00 GMT

Overview

Auto-generated Go protobuf bindings defining the core ML Metadata data model types including artifacts, executions, contexts, events, and database configuration structures.

Description

Generated by `protoc-gen-go` (v1.26.0) from `ml_metadata/proto/metadata_store.proto`, this file (6,422 lines) defines Go struct types and methods for all MLMD domain objects. Key types include: Artifact (with states: PENDING, LIVE, MARKED_FOR_DELETION, DELETED, ABANDONED, REFERENCE), Execution, Context, Event (linking artifacts to executions), Attribution (linking artifacts to contexts), Association (linking executions to contexts), LineageGraph, Value (variant container for int/double/string/struct/proto/bool), and various database configuration types (MySQL, SQLite, PostgreSQL).

Usage

Imported by the KFP Go backend to interact with the ML Metadata store for pipeline artifact tracking, execution history, and lineage queries.

Code Reference

Source Location

• Repository: Kubeflow_Pipelines
• File: third_party/ml-metadata/go/ml_metadata/metadata_store.pb.go
• Lines: 1-6422

Signature

package ml_metadata

type Artifact struct {
    Id         *int64
    TypeId     *int64
    Uri        *string
    State      *Artifact_State
    Properties map[string]*Value
    // ...
}

type Execution struct {
    Id         *int64
    TypeId     *int64
    Properties map[string]*Value
    // ...
}

type Context struct {
    Id         *int64
    TypeId     *int64
    Name       *string
    Properties map[string]*Value
    // ...
}

type Event struct {
    ArtifactId  *int64
    ExecutionId *int64
    Type        *Event_Type  // INPUT, OUTPUT, etc.
    Path        *Event_Path
}

Import

import ml_metadata "github.com/kubeflow/pipelines/third_party/ml-metadata/go/ml_metadata"

I/O Contract

Inputs

Name	Type	Required	Description
Proto descriptor	[]byte	Yes	Raw protobuf descriptor from metadata_store.proto

Outputs

Name	Type	Description
Domain types	Go structs	Artifact, Execution, Context, Event, Value, LineageGraph, etc.
Config types	Go structs	ConnectionConfig, MySQLDatabaseConfig, SqliteMetadataSourceConfig, etc.

Usage Examples

Creating an Artifact

import ml_metadata "github.com/kubeflow/pipelines/third_party/ml-metadata/go/ml_metadata"

artifact := &ml_metadata.Artifact{
    Uri:    proto.String("gs://bucket/model/output"),
    TypeId: proto.Int64(typeId),
    Properties: map[string]*ml_metadata.Value{
        "name": {Value: &ml_metadata.Value_StringValue{StringValue: "trained-model"}},
    },
}

Related Pages

• Environment:Kubeflow_Pipelines_KFP_Backend_Deployment