Skip to content

Security: AKIVA-AI/toolkit-llm-gateway

Security

SECURITY.md

Security Policy

Supported Versions

Version Supported
latest on main Yes
older releases Best effort

Reporting a Vulnerability

Do not open a public issue for security vulnerabilities.

Preferred: GitHub Security Advisory

  1. Navigate to the repository's Security tab.
  2. Click Report a vulnerability under "Private vulnerability reporting".
  3. Fill in the details including affected version, reproduction steps, and impact assessment.

Alternative: Email

If private advisory reporting is unavailable, email the maintainer directly with:

  • Affected version or commit SHA
  • Reproduction steps or proof-of-concept
  • Impact assessment (confidentiality, integrity, availability)
  • Any suggested fix (optional)

Response Timeline

  • Acknowledgement: within 3 business days of receipt.
  • Triage and severity assessment: within 7 business days.
  • Fix and coordinated disclosure: target 30 days from acknowledgement, negotiable for complex issues.

Security Measures

This project enforces the following security practices:

  • Dependency scanning: Dependabot monitors pip and GitHub Actions dependencies weekly.
  • Static analysis: Bandit (SAST) runs in CI and blocks merges on findings.
  • Dependency CVE checks: Safety scans run in CI and block merges on known vulnerabilities.
  • Constant-time key comparison: API key verification uses hmac.compare_digest.
  • No query-parameter credentials: API keys are only accepted via the X-API-Key header.
  • No default passwords: Docker Compose requires explicit POSTGRES_PASSWORD via environment.
  • Non-root containers: Docker images run as a non-root user.

Scope

The following are in scope for security reports:

  • Any code in src/toolkit_extensions/ and dashboard/
  • Docker and deployment configurations
  • CI/CD pipeline security (workflow injection, secret leakage)
  • Dependency vulnerabilities in direct dependencies

The forked LiteLLM code in src/ (outside toolkit_extensions/) should be reported upstream to the LiteLLM project.

There aren’t any published security advisories