Java

InferenceKey ships native bindings today for Python (inferencekey) and Node / TypeScript (@inferencekey/sdk). A Java client is on the roadmap.

Coming soon

The Java SDK is not released yet. It is planned as a thin binding over the same inferencekey-core Rust library that powers every other language, exposed through the C ABI (via the Foreign Function & Memory API, with a JNI fallback for older JDKs).

What to expect

• Distribution. Published to Maven Central under the intended coordinates dev.inferencekey:inferencekey-sdk.
• Same surface. The API will mirror the other languages: a ManagementClient for the control plane (ik_sdk_ tokens) and a DataClient for the data plane (ik_live_ tokens), with ensure(...) for idempotent provisioning and endpoint(...) for inference. Feature parity is reached over time — newer conveniences such as streaming and readiness helpers may land after the first release.
• Same model. Two tokens, two clients, least privilege. The control client provisions and reconciles workloads; the data client calls inference per workload. See Tokens for the split.
• Same configuration. The shared environment variables apply: INFERENCEKEY_BASE_URL, INFERENCEKEY_PROJECT, INFERENCEKEY_API_KEY, and INFERENCEKEY_SDK_TOKEN.

In the meantime

You can use any of the shipped languages today, or call the wire API directly (OpenAI-compatible data plane under /endpoint/:projectSlug/:workloadSlug/v1/...).

Python SDK Shipped today: pip install inferencekey.

Node / TypeScript SDK Shipped today: @inferencekey/sdk.

---

New to InferenceKey? Create an account or open the dashboard · Learn more at inferencekey.com.