Skip to content

CONTRIBUTING.md

Latest commit

 

History

History
54 lines (39 loc) · 1.39 KB

CONTRIBUTING.md

File metadata and controls

54 lines (39 loc) · 1.39 KB
Raw

Contributing

Thanks for contributing to semblend-vllm-connector.

This repo is an out-of-tree vLLM connector. Keep changes small, reviewable, and consistent with vLLM connector lifecycle semantics.

Development Setup

python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"
make check

With SemBlend provider support:

pip install -e ".[semblend,dev]"

Pull Requests

Before opening a PR:

  • run make check;
  • add or update tests for behavior changes;
  • update docs for user-visible behavior changes;
  • keep discovery and materialization behavior clearly separated;
  • keep the repo focused on the vLLM connector surface.

Use one of these PR title prefixes when practical:

  • [Bugfix]
  • [Connector]
  • [Config]
  • [Docs]
  • [Test]
  • [CI]

vLLM Connector Rules

  • Do not weaken exact prefix-cache semantics.
  • Do not return positive matched tokens unless KV can actually be loaded.
  • Do not publish non-identical semantic donor KV into the exact prefix cache.
  • Validate model, tokenizer, namespace, cache salt, adapter, and block shape.
  • Fail closed to normal vLLM prefill on unsupported cases.

Contribution Rights

By submitting a contribution, you agree that your contribution is provided under the Apache-2.0 license used by this repository and that you have the right to submit it. Maintainers may request a Signed-off-by line for larger changes.