github.devotd.
ReleaseSecurity

archestra-ai/openappa v0.35.1 — publish policy after head check

archestra-ai/openappa v0.35.1 is a bug fix introducing Runtime::publish_if_head, which makes the runtime reread its input when the head has moved.

Repositoryarchestra-ai/openappa ↗

Deterministic guardrails that don't break agents

stars
★ 1.6K
Forks
67
Language
Rust
License
MIT
Latest release
v0.35.1
Read original source ↗View on GitHub ↗

archestra-ai/openappa v0.35.1 is a bug fix that introduces Runtime::publish_if_head, letting an embedded host publish a policy for the head it observed before validation. According to the GitHub release notes, if the head observed by the host has already moved, this method returns PolicyHeadMoved and forces the host to reread its input. The core of this change is handling head movement on the path that publishes a policy based on the observed head. When the head has changed, the flow goes back to rereading the input instead of using the existing input as-is. This is a fix that addresses the case where the head has moved, rather than the addition of a new policy feature.

This repository aims at deterministic guardrails that do not break agents. This release contains the change corresponding to checking the head at the time the policy is published. Runtime::publish_if_head lets an embedded host publish a policy for the head it observed before validation. What the release notes disclose, however, is only the behavior of the method, the return value on head movement, and the input reread. Since it is not stated which head the head concretely refers to, or what behavior the policy restricts, the scope cannot be described any more broadly than that.

There is no schema change. This means this release does not include a schema change, and it is not a basis for judging whether other adjustments are needed. When upgrading from a version before v0.35.0, the existing upgrade procedure is still required. The release notes do not restate the procedure itself, so no extra steps can be added. Therefore the confirmed scope is a bug fix that rereads the input on head movement, no schema change, and the continuation of the existing upgrade procedure. No information about performance or adoption impact is given, so evaluation of this change is limited to the documented behavior.