Implementation:Kubeflow Pipelines MLMD Service GRPC

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

Overview

Auto-generated Go gRPC client and server stubs for the ML Metadata Store service, providing the RPC interface for all metadata CRUD operations.

Description

Generated by `protoc-gen-go-grpc` (requiring gRPC-Go v1.32.0+), this file (2,441 lines) defines the MetadataStoreServiceClient interface with 50+ RPC methods, a concrete `metadataStoreServiceClient` implementation using `grpc.ClientConn`, and the MetadataStoreServiceServer interface for server implementations. Also includes `UnimplementedMetadataStoreServiceServer` for forward-compatible servers.

Usage

The KFP API server uses the client stubs from this file to record and query pipeline metadata. This completes the gRPC integration layer alongside the message types from the companion `_service.pb.go` file.

Code Reference

Source Location

• Repository: Kubeflow_Pipelines
• File: third_party/ml-metadata/go/ml_metadata/metadata_store_service_grpc.pb.go
• Lines: 1-2441

Signature

package ml_metadata

type MetadataStoreServiceClient interface {
    PutArtifactType(ctx context.Context, in *PutArtifactTypeRequest, opts ...grpc.CallOption) (*PutArtifactTypeResponse, error)
    PutArtifacts(ctx context.Context, in *PutArtifactsRequest, opts ...grpc.CallOption) (*PutArtifactsResponse, error)
    PutExecutions(ctx context.Context, in *PutExecutionsRequest, opts ...grpc.CallOption) (*PutExecutionsResponse, error)
    GetArtifactsByType(ctx context.Context, in *GetArtifactsByTypeRequest, opts ...grpc.CallOption) (*GetArtifactsByTypeResponse, error)
    GetLineageGraph(ctx context.Context, in *GetLineageGraphRequest, opts ...grpc.CallOption) (*GetLineageGraphResponse, error)
    // ... 50+ RPC methods
}

func NewMetadataStoreServiceClient(cc grpc.ClientConnInterface) MetadataStoreServiceClient

type MetadataStoreServiceServer interface {
    // Same methods as client, but (ctx, *Request) -> (*Response, error)
}

type UnimplementedMetadataStoreServiceServer struct{}

Import

import ml_metadata "github.com/kubeflow/pipelines/third_party/ml-metadata/go/ml_metadata"
import "google.golang.org/grpc"

I/O Contract

Inputs

Name	Type	Required	Description
grpc.ClientConnInterface	gRPC connection	Yes	Established gRPC connection to MLMD server

Outputs

Name	Type	Description
MetadataStoreServiceClient	interface	Client stub for making MLMD gRPC calls
MetadataStoreServiceServer	interface	Server interface for implementing MLMD service

Usage Examples

Creating a Client and Putting Artifacts

import (
    ml_metadata "github.com/kubeflow/pipelines/third_party/ml-metadata/go/ml_metadata"
    "google.golang.org/grpc"
)

conn, err := grpc.Dial("metadata-grpc-service:8080", grpc.WithInsecure())
if err != nil {
    log.Fatalf("Failed to connect: %v", err)
}
defer conn.Close()

client := ml_metadata.NewMetadataStoreServiceClient(conn)
resp, err := client.PutArtifacts(ctx, &ml_metadata.PutArtifactsRequest{
    Artifacts: []*ml_metadata.Artifact{
        {Uri: proto.String("gs://bucket/output"), TypeId: proto.Int64(1)},
    },
})

Related Pages

• Environment:Kubeflow_Pipelines_KFP_Backend_Deployment