Implementation:SeldonIO Seldon core Seldon Pipeline Infer

Field	Value
Implementation Name	Seldon Pipeline Infer
Type	External Tool Doc
Overview	Concrete CLI tool for sending V2-protocol inference requests to Seldon Core 2 pipelines.
Related Principle	SeldonIO_Seldon_core_Pipeline_Inference_Execution
Source	docs-gb/cli/seldon_pipeline_infer.md:L1-35
Domains	MLOps, Inference
External Dependencies	seldon CLI, curl, grpcurl, V2 protocol, Kafka
Knowledge Sources	Repo (https://github.com/SeldonIO/seldon-core), Doc (https://docs.seldon.io/projects/seldon-core/en/v2/)
Last Updated	2026-02-13 00:00 GMT

Description

The seldon pipeline infer command sends a V2 Inference Protocol request to a deployed Seldon Core 2 pipeline and returns the prediction output. It supports both REST and gRPC inference modes, repeated iterations for benchmarking, and custom headers. The data flows through the pipeline's DAG steps via Kafka, and the response contains the output tensors from the designated output steps.

Code Reference

CLI Signature

seldon pipeline infer <pipelineName> (data) [flags]

CLI Options

Flag	Description	Default
pipelineName	Name of the pipeline to call (positional)	(required)
data	V2 JSON payload (positional)	(required unless -f used)
-f, --file-path	Inference payload file	(none)
--inference-mode	Inference mode: rest or grpc	rest
--inference-host	Seldon inference host	0.0.0.0:9000
-i, --iterations	How many times to run inference	1
-t, --seconds	Number of seconds to run inference	(none)
--header	Add a header (key=value); repeatable	(none)
--show-headers	Show request and response headers	false
-r, --show-request	Show the request payload	false
-o, --show-response	Show the response payload	true
-s, --sticky-session	Use sticky session from last inference	false

Source: docs-gb/cli/seldon_pipeline_infer.md:L1-35

I/O Contract

Inputs

• V2 JSON payload: A JSON object following the V2 Inference Protocol schema:

{
  "inputs": [
    {
      "name": "INPUT0",
      "data": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
      "datatype": "INT32",
      "shape": [1, 16]
    },
    {
      "name": "INPUT1",
      "data": [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16],
      "datatype": "INT32",
      "shape": [1, 16]
    }
  ]
}

Each input tensor requires:

• name (string): Tensor name matching the first step's expected input.
• data (list): Flattened tensor data values.
• datatype (string): Data type (e.g., INT32, FP32, BYTES).
• shape (list of int): Tensor dimensions.

Outputs

• V2 JSON response with outputs from the final pipeline step(s):

{
  "model_name": "",
  "outputs": [
    {
      "data": [2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32],
      "name": "OUTPUT0",
      "shape": [1, 16],
      "datatype": "INT32"
    },
    {
      "data": [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
      "name": "OUTPUT1",
      "shape": [1, 16],
      "datatype": "INT32"
    }
  ]
}

Usage Examples

Basic Pipeline Inference

seldon pipeline infer tfsimples \
    '{"inputs":[{"name":"INPUT0","data":[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16],"datatype":"INT32","shape":[1,16]},{"name":"INPUT1","data":[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16],"datatype":"INT32","shape":[1,16]}]}' | jq -M .

Inference via REST (curl)

curl -X POST http://0.0.0.0:9000/v2/pipelines/tfsimples/infer \
  -H "Content-Type: application/json" \
  -d '{"inputs":[{"name":"INPUT0","data":[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16],"datatype":"INT32","shape":[1,16]},{"name":"INPUT1","data":[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16],"datatype":"INT32","shape":[1,16]}]}' | jq -M .

Inference via gRPC

seldon pipeline infer tfsimples \
    '{"inputs":[{"name":"INPUT0","data":[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16],"datatype":"INT32","shape":[1,16]},{"name":"INPUT1","data":[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16],"datatype":"INT32","shape":[1,16]}]}' \
    --inference-mode grpc

Repeated Inference for Benchmarking

seldon pipeline infer tfsimples \
    '{"inputs":[{"name":"INPUT0","data":[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16],"datatype":"INT32","shape":[1,16]},{"name":"INPUT1","data":[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16],"datatype":"INT32","shape":[1,16]}]}' \
    -i 100

Inspecting Pipeline Data Flow

# Trace data through all pipeline steps for debugging
seldon pipeline inspect tfsimples

Inference from a File

seldon pipeline infer tfsimples -f ./payload.json

Related Pages

• SeldonIO_Seldon_core_Pipeline_Inference_Execution - implements - Principle of dataflow-based inference through multi-step pipelines.
• SeldonIO_Seldon_core_Seldon_Pipeline_Status - prerequisite - Pipeline must be confirmed ready before inference.
• SeldonIO_Seldon_core_Seldon_Pipeline_CRD - defines topology - The pipeline CRD determines how the inference request flows through steps.
• SeldonIO_Seldon_core_Seldon_Pipeline_Advanced_Routing - routing logic - Conditional routing patterns that affect inference flow.
• Environment:SeldonIO_Seldon_core_Kubernetes_Cluster_Environment
• Environment:SeldonIO_Seldon_core_Docker_Compose_Local_Environment
• Environment:SeldonIO_Seldon_core_Kafka_Messaging_Environment
• Heuristic:SeldonIO_Seldon_core_Kafka_Partition_Throughput_Tip