coralapi

Security Policy

Supported Versions

Version	Supported
latest	Yes

Reporting a Vulnerability

If you discover a security vulnerability, please report it responsibly. For issues that do not require private disclosure, please open a GitHub issue.

For sensitive security issues that require private disclosure, please contact the maintainer directly.

Please include:

• A description of the vulnerability
• Steps to reproduce the issue
• Any potential impact
• Suggested fix (if you have one)

You should receive a response within 72 hours. Security fixes will be prioritized and released as soon as possible.

Known Limitations

CoralAPI ships without authentication or rate limiting by design — it is meant to run behind an authenticated gateway or ingress that provides those. See the Security model section of the README for the full posture. In summary:

• No auth / no rate limiting: the trust boundary is a fronting gateway. Do not expose the service directly to untrusted networks.
• Async job IDs are bearer secrets: job results are readable by anyone holding the (unguessable, UUID4) job ID; there is no per-caller ownership.
• Model downloads run server-side against a configured HTTPS source with redirects disabled, per-model and total-cache byte caps, and optional sha256 pinning (CORALAPI_MODEL_CHECKSUMS). Point CORALAPI_MODEL_SOURCE only at a source you trust; models are loaded onto the TPU.
• Uploads are streamed to disk with a hard size limit enforced at the ASGI layer and a maximum image resolution; tune CORALAPI_MAX_UPLOAD_BYTES and CORALAPI_MAX_IMAGE_DIM for your environment.

This site is open source. Improve this page.