Skip to content

Security: unison-labs-ai/python-sdk

Security

SECURITY.md

Security Policy

Reporting a vulnerability

Please report security issues privately — do not open a public GitHub issue.

Email security@unisonlabs.ai with:

  • a description of the issue and its impact,
  • steps to reproduce (a proof-of-concept if you have one),
  • any suggested remediation.

We aim to acknowledge within 3 business days and to keep you updated as we investigate. We will credit reporters who want it once a fix ships.

Scope

This repository is the Python SDK for the Unison brain API. It holds no secrets and is not a security boundary — all authentication, authorization, workspace isolation, and rate limiting are enforced server-side by the Unison brain API. Reports about the client are most useful when they concern:

  • credential handling in environment variables or on disk,
  • dependency or supply-chain risks (e.g. malicious transitive deps),
  • accidental credential logging or exposure via error messages.

Server-side or account issues should also go to the same address.

Handling of credentials

The client reads your API key from the UNISON_TOKEN environment variable (or the token= constructor argument). The token is never written to disk by this library, never logged, and is transmitted only to the configured API host (UNISON_API_URL, default https://brain.unisonlabs.ai) as an HTTP Authorization: Bearer <token> header.

Store your key in a .env file (covered by .gitignore) or a secrets manager. Never hard-code it in source files.

There aren't any published security advisories