Repository files navigation

Gateway API Inference Extension optimizes self-hosting Generative Models on Kubernetes.This is achieved by leveraging Envoy'sExternal Processing (ext-proc) to extend any gateway that supports both ext-proc andGateway API into aninference gateway.

Inference Gateway has partnered with vLLM to accelerate LLM serving optimizations withllm-d!

The following specific terms to this project:

• Inference Gateway (IGW): A proxy/load-balancer which has been coupled with anEndpoint Picker. It provides optimized routing and load balancing forserving Kubernetes self-hosted generative Artificial Intelligence (AI)workloads. It simplifies the deployment, management, and observability of AIinference workloads.
• Inference Scheduler: An extendable component that makes decisions about which endpoint is optimal (best cost /best performance) for an inference request based onMetrics and CapabilitiesfromModel Serving.
• Metrics and Capabilities: Data provided by model serving platforms aboutperformance, availability and capabilities to optimize routing. Includesthings likePrefix Cache status orLoRA Adapters availability.
• Endpoint Picker(EPP): An implementation of anInference Scheduler with additional Routing, Flow, and Request Control layers to allow for sophisticated routing strategies. Additional info on the architecture of the EPPhere.

The following are key industry terms that are important to understand forthis project:

• Model: A generative AI model that has learned patterns from data and isused for inference. Models vary in size and architecture, from smallerdomain-specific models to massive multi-billion parameter neural networks thatare optimized for diverse language tasks.
• Inference: The process of running a generative AI model, such as a largelanguage model, diffusion model etc, to generate text, embeddings, or otheroutputs from input data.
• Model server: A service (in our case, containerized) responsible forreceiving inference requests and returning predictions from a model.
• Accelerator: specialized hardware, such as Graphics Processing Units(GPUs) that can be attached to Kubernetes nodes to speed up computations,particularly for training and inference tasks.

For deeper insights and more advanced concepts, refer to ourproposals.

This extension upgrades anext-proc capable proxy or gateway - such as Envoy Gateway, kGateway, or the GKE Gateway - to become aninference gateway - supporting inference platform teams self-hosting Generative Models (with a current focus on large language models) on Kubernetes. This integration makes it easy to expose and control access to your localOpenAI-compatible chat completion endpoints to other workloads on or off cluster, or to integrate your self-hosted models alongside model-as-a-service providers in a higher levelAI Gateway like LiteLLM, Solo AI Gateway, or Apigee.

The Inference Gateway:

• Improves the tail latency and throughput of LLM completion requests against Kubernetes-hosted model servers using an extensible request scheduling alogrithm that is kv-cache and request cost aware, avoiding evictions or queueing as load increases
• ProvidesKubernetes-native declarative APIs to route client model names to use-case specific LoRA adapters and control incremental rollout of new adapter versions, A/B traffic splitting, and safe blue-green base model and model server upgrades
• Adds end to end observability around service objective attainment
• Ensures operational guardrails between different client model names, allowing a platform team to safely serve many different GenAI workloads on the same pool of shared foundation model servers for higher utilization and fewer required accelerators

IGW’s pluggable architecture was leveraged to enable thellm-d Inference Scheduler.

Llm-d customizes vLLM & IGW to create a disaggregated serving solution. We've worked closely with this team to enable this integration. IGW will continue to work closely with llm-d to generalize the disaggregated serving plugin(s), & set a standard for disaggregated serving to be used across anyprotocol-adherent model server.

IGW has enhanced support for vLLM via llm-d, and broad support for any model servers implementing the protocol. More details can be found inmodel server integration.

This project is in alpha. latest release can be foundhere.
It should not be used in production yet.

Follow ourGetting Started Guide to get the inference-extension up and running on your cluster!

Seeour website for detailed API documentation on leveraging our Kubernetes-native declarative APIs

As Inference Gateway builds towards a GA release. We will continue to expand our capabilities, namely:

1. Prefix-cache aware load balancing with interfaces for remote caches
2. Recommended LoRA adapter pipeline for automated rollout
3. Fairness and priority between workloads within the same criticality band
4. HPA support for autoscaling on aggregate metrics derived from the load balancer
5. Support for large multi-modal inputs and outputs
6. Support for other GenAI model types (diffusion and other non-completion protocols)
7. Heterogeneous accelerators - serve workloads on multiple types of accelerator using latency and request cost-aware load balancing
8. Disaggregated serving support with independently scaling pools

Follow thisREADME to learn more about running the inference-extension end-to-end test suite on your cluster.

Our community meeting is weekly at Thursday 10AM PDT (Zoom,Meeting Notes).

We currently utilize the#gateway-api-inference-extension channel in Kubernetes Slack workspace for communications.

Contributions are readily welcomed, follow thedev guide to start contributing!

Participation in the Kubernetes community is governed by theKubernetes Code of Conduct.