2.0.2

• chore(deps): update runpod to latest version

2.0.1

• fix: allow more cuda versions; use proper name of the MODEL_NAME

2.0.0

• feat: added tool_call_parser; updated sglang to 0.5.2
• feat: use "MODEL_NAME" instead of "MODEL_PATH"

breaking changes

• MODEL_PATH → MODEL_NAME to use the same naming as worker-vllm
• CUDA 12.6 minimum (was 12.4)

1.2.0

What's Changed

• feat: enable tool calling support by @DifferentialityDevelopment in #25

New Contributors

• @DifferentialityDevelopment made their first contribution in #25

1.1.2

• docs: added better examples for users to use sglang

1.1.1

• fix: use proper type for HF_TOKEN

1.1.0

• feat: added HF_TOKEN support in the runpod hub

1.0.0

This release adds this worker into the hub and is from now on using the official docker images from sglang instead of installing sglang manually.

What's Changed

• update worker-config by @pandyamarut in #15
• fix dynamic batching by @pandyamarut in #17
• fix: E-2702: remove broken CI by @ef0xa in #19
• feat: add sglang to the hub by @TimPietrusky in #23

New Contributors

• @ef0xa made their first contribution in #19
• @TimPietrusky made their first contribution in #23

v0.2.0

• Sglang Version update from 0.3.3-> v0.4.1.post3

v0.1.0

Initial Sglang worker release.

• Pre-built docker images.
• Engine configuration through the Environment variables.