Security policy

Reporting a vulnerability

Please do not open a public issue for a security problem.

Use GitHub's private vulnerability reporting: go to the Security tab of this repository and choose Report a vulnerability. That opens a private advisory visible only to you and the maintainers. If you cannot use that form, write to support@oriveoai.com with SECURITY in the subject.

Please include what you can of:

• which client is affected (iOS, Android, web) and which version
• what an attacker can do, and what they need in order to do it
• the steps to reproduce it
• anything you already know about the fix

You will get an acknowledgement within a few days. Please give us a reasonable window to ship a fix before disclosing publicly; we will credit you in the advisory unless you would rather we did not.

What this project treats as a vulnerability

This is a bring-your-own-key client. Its security promises are narrow and specific, and these are the ones worth reporting against:

• Credential exposure. A provider API key reaching anywhere other than the provider it belongs to — written to a log, included in an error report, sent to the wrong endpoint, left in an export that was supposed to exclude it, or readable by another app on the device.
• Request forgery. Anything that makes a client issue a request to a host the user did not configure. The relay forwarder in the web client resolves a hostname once and pins the resulting address for the connection, refuses loopback, private and link-local ranges, accepts only a short list of ports, and re-checks every redirect hop against the same rules. A way around any of that is a vulnerability. Two things are deliberate and are not: the address pin is applied in a production build, and the carrier-grade NAT range 100.64.0.0/10 is permitted on purpose, so a host reached over a mesh VPN stays reachable.
• Data at rest. Reading another storage partition's conversations or keys, or defeating the encryption on a password-protected backup archive.
• Untrusted content escaping its frame. Model output, a note, or an attachment causing code execution, or reaching a capability the user did not grant.
• Supply chain. A dependency in this repository with a known exploitable vulnerability that this project actually reaches.

What is out of scope

• Prompt injection and model behaviour. A model that can be talked into saying something is a property of the model, not a defect in this client. Reports that a model produced unwanted output belong with that model's provider.
• A key that a user pasted into a relay they chose to trust. Configuring a third-party endpoint means giving it your key; that is the nature of the feature. A bug in how the key is sent is in scope; the decision to send it is not.
• Cleartext HTTP to a model server on your own network. This is deliberate and documented — a local engine on a private address typically has no certificate. Reaching a public host in cleartext would be a bug; reaching your own machine is the feature.
• The proprietary Oriveo apps on the App Store, Google Play, and the hosted web app. Those are a separate product; report issues with them through support@oriveoai.com rather than here.
• Missing hardening headers, or the absence of a defence, with no demonstrated impact.

Supported versions

This project is developed on main, and fixes land there. There are no long-lived release branches to backport to.