OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Infrastructure & Serving

Openlake

OpenLake is an open-source, Rust-based distributed storage engine for GPU workloads, including LLM inference and training. Its repository documents a vLLM connector for KV-cache storage and describes additional uses such as checkpointing, vector workloads, and agent-context retrieval.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Openlake — OpenViST project profile in AI Infrastructure & Serving.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Consider OpenLake when building or adapting GPU-backed inference or training infrastructure and the need is a storage layer, rather than a model-serving application. The repository's most concrete setup path connects its KV store to vLLM.

Who it is for

  • Teams operating vLLM inference on GPU hosts that want to configure OpenLake as KV-cache storage.
  • AI infrastructure engineers evaluating persistent storage for training, checkpointing, vector, or context-retrieval workloads.

Documented uses

  • Configure vLLM to write and read inference KV cache through an OpenLake store.
  • Share a KV pool across GPU hosts using configured OpenLake nodes and ordered vLLM peer addresses.
  • Evaluate a storage engine for training reads, checkpointing, vector workloads, or agent context.
Before you choose

Requirements & limits

The project repository is public and identifies the software as Apache-2.0 licensed. The README shows a Linux-oriented build path and identifies Rust 1.91 or newer.

  • The documented inference integration is specifically a vLLM connector configuration.
  • The quickstart uses same-host offload by default; cross-host use requires store configuration and ordered node addresses.
  • The implementation is identified as Rust-based and built on io_uring.
  • The README's throughput and latency statements are project claims rather than independent performance evidence.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Configure OpenLake KV storage for vLLM on one host

Start the OpenLake store and configure vLLM to use its connector for KV transfer on the same host.

Before you start

  • A Linux GPU host with vLLM available.
  • Python package installation and the OpenLake command openlaked.
  • A model name for the vLLM serving command.

Steps

  1. Install the connector with the command below.

    Documented command
    pip install openlake-vllm

  2. Start the store with openlaked.

  3. Set PYTHONHASHSEED=0 and run vllm serve with the README's --kv-transfer-config JSON, using OpenLakeConnector, openlake_client.openlake_connector, kv_role kv_both, node 127.0.0.1:9400, and device local.

Expected outcome

The documented vLLM configuration points its KV connector at the OpenLake store on the same host.

Guide limitations

  • The quickstart defaults to same-host offload; a GPU fleet requires starting openlaked with a configuration.
  • This recipe covers the documented configuration steps, not a tested deployment or performance result.
Documentation 1 ↗
COMPARE OPTIONS

Compare options for Model inference and serving

These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.

ProjectDocumented purposePlatformAccess and requirements
OpenlakeConfigure vLLM to write and read inference KV cache through an OpenLake store.Linux, ServerThe project repository is public and identifies the software as Apache-2.0 licensed. The README shows a Linux-oriented build path and identifies Rust 1.91 or newer.
MorphingdbImport a libtorch model and create predictions from values in a PostgreSQL table.Docker, Linux, Self-hosted, ServerThe README documents a Docker image and a build-from-repository option, followed by connecting to PostgreSQL and enabling the pgdl extension.
Gateway — pyadavSend chat-completion requests to a provider through a common server endpoint.ServerThe repository is public and identifies the project as Apache-2.0 licensed. README installation instructions include binaries, platform package managers and Docker.
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
11 October 2026
Original website title
openlake-project/openlake · Official homepage title · retained in its source language
How we review sourcesSuggest a correction