Skip to content

Security: brtydse100/vllm-optimizer

Security

SECURITY.md

Security policy

Configuration and reports recursively redact conventionally named credentials, including tokens, passwords, authorization headers, cookies, access keys, and client secrets. Command redaction covers --key value and --key=value forms. Redaction changes display and persistence only, never subprocess input. Arbitrary secrets stored under non-sensitive names cannot be recognized reliably.

Supported versions

Security fixes are applied to the latest published release. Alpha releases are experimental and should not be used with untrusted configuration files or model-serving endpoints.

Reporting a vulnerability

Please report vulnerabilities privately through the repository's GitHub security-advisory process. Do not include API keys, model credentials, or benchmark data in a public issue. Include the affected version, a minimal reproduction, and the impact when safe to do so.

Reports are acknowledged within five business days. We will coordinate a fix, release, and disclosure timeline with the reporter.

There aren't any published security advisories