What Seldon Core 2 does
Seldon Core 2 is an open-source Kubernetes model-serving platform with reusable servers, model scheduling, pipelines, dataflow, metrics, and deployment APIs.
Seldon Core 2 focuses on efficient model operations at cluster scale. Models can be loaded onto shared servers, updated independently, and connected into synchronous or dataflow pipelines. It supports local Docker Compose or Kind learning environments and production Kubernetes, but the dependency footprint grows when pipelines, gateways, Kafka, and observability are enabled.
The open-source platform is free. Kubernetes nodes, accelerators, Kafka for dataflow pipelines, Envoy, storage, network, monitoring, upgrades, and engineering determine total cost; commercial Seldon offerings or support are separate. Install only necessary gateways and dataflow components to reduce capacity and attack surface.
Seldon documents TLS options for control plane, Kafka, and data plane, yet secure values and certificates must be configured deliberately. Plaintext defaults or broad Kubernetes access can expose models and data. Use namespace isolation, RBAC, network policies, encrypted transport, protected artifact stores, image scanning, audit logs, model validation, and human oversight for consequential predictions.
How Seldon Core 2 works
Operators install Seldon Core 2 custom resources and runtime components, then declare servers, models, experiments, or pipelines. The control plane schedules model artifacts onto compatible reusable servers; gateways route inference requests; optional Kafka-backed dataflow coordinates multi-step pipelines. Model runtimes compute predictions and return responses while metrics and tracing expose operations. The organization owns cluster and transport security, artifact access, runtime compatibility, data governance, and human review.
How to set up Seldon Core 2
Choose a minimal topology
Determine model types, server reuse, synchronous or dataflow pipelines, Kubernetes version, gateways, Kafka need, throughput, and availability targets.
Install in a test environment
Pin compatible Helm charts and dependencies, start with only required components, and validate the learning or namespace-scoped configuration.
Secure every plane
Configure certificates and TLS for control, data, and Kafka paths; apply RBAC, network policy, secret management, and private artifact access.
Deploy representative models
Pin artifacts and server images, declare resources and replicas, verify input schemas, and benchmark loading, routing, prediction quality, and failure recovery.
Operate and upgrade
Monitor metrics and traces, budget capacity, test certificate rotation and rollback, stage version changes, and require review for high-impact predictions.
Seldon Core 2 FAQs
Is Seldon Core 2 free?
Yes, the core software is open source. Kubernetes, compute, storage, Kafka, observability, operations, and optional commercial services are separate costs.
Does it require Kubernetes?
Production guidance targets Kubernetes. Docker Compose and Kind options are available for local learning and testing.
Is Kafka always required?
No. Documentation says Kafka is required for dataflow pipelines; teams not using them can omit the related gateways and engine.
Can it encrypt traffic?
It supports secure control, Kafka, and data-plane configurations, but administrators must provide certificates, choose secure protocols, and verify enforcement.
Does it validate model predictions?
Operational metrics do not prove prediction quality. Teams must evaluate drift, fairness, errors, uncertainty, and human-review requirements separately.
Listing reviewed 2026-08-03. Product details and pricing can change; verify important terms on the provider's website.
Related Developer Tools AI tools
Related AI guides
Reviews
Tell the community what you made, what worked, and what you wish you knew before starting.