Why is a normal OpenAPI diff not enough for my APIs?

Structural diffs catch removed operations and narrowed schemas, but miss prose changes to pagination, ordering, retries, auth, and error meaning. This layer asks Jev bounded semantic questions only about changed contract text.

Will fuzzy model output start blocking my releases?

Advisory mode is the default: structural breaks block while semantic risks request review. Enforcement needs strong probabilities on multiple heads, and review findings can optionally fail the run.

How do I keep keys and contract text safe in CI?

Keys stay in environment or key files and never in committed config, remote references are rejected rather than fetched, local references stay jailed, and traces redact credentials with owner-only file permissions.

Can I quiet legacy endpoints without hiding new breaks?

Yes. Suppressions need an owner, reason, expiry, and pattern; matches become notices with original severity preserved, and expired suppressions fail the run instead of silently persisting.

How do I install and run the sentinel CLI without adding it to my project?

Run it once with uvx jev-oas-sentinel --version, or install it isolated with uv tool install jev-oas-sentinel or pipx install jev-oas-sentinel. Without uv it needs Python 3.9+. The published package is on PyPI and the code is Apache-2.0 licensed.

Which compatibility checks run without ever calling Jev?

Pass --no-jev and the deterministic layer still catches removed operations, removed or newly required parameters, new required request bodies, removed status codes and media types, removed or non-guaranteed properties, enum narrowing and expansion, incompatible type or nullability changes, and security requirement changes.

What does enforce mode change compared with the default?

Advisory mode is the default: definite structural breaks block and semantic risks request review. Enforce only blocks a semantic break when both the breaking probability and the promise-violation probability cross conservative thresholds. --fail-on-review returns a non-zero exit code for review findings too.

CommunityJev Agents & MCP

Jev Oas Sentinel

OpenAPI compatibility gate that pairs deterministic structural checks with bounded Jev questions over changed prose to catch behavioral breaks that schema diffs miss.

See highlights

Open github.com

Stars: 5
Language: Python
License: APACHE-2.0
Published by: Community
Author: ShuhanSun
Last verified: Oct 2026

ShuhanSun/jev-oas-sentinel

User repository on GitHub

View on GitHub

Catch breaking API behavior hidden in OpenAPI prose with deterministic checks and TypeSafe JEV System One semantic review.

Forks: 1
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

About this project

Catch behavioral breaking changes hidden in OpenAPI prose—changes that structural schema diff tools cannot see. jev-oas-sentinel compares two OpenAPI documents in two layers: deterministic checks find definite structural compatibility problems; TypeSafe Jev evaluates bounded semantic questions about changed descriptions, examples, defaults, retry behavior, ordering, pagination, authorization, and error semantics.

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Python

Highlights

Runs deterministic checks first, then asks Jev only about changed descriptions, examples, defaults, and semantics.
Dry-run and no-model modes preview semantic review without a key; live runs cap calls, timeouts, and retries.
Policy file sets advisory or enforce modes with review and block thresholds plus expiring suppressions.
Emits Markdown, JSON, and SARIF with sanitized opt-in traces and native certificate verification.

Watch out

Semantic verdicts are probabilistic, so the digest advises against enforcement mode until questions and thresholds are evaluated on your own APIs. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

jev-recruiter

Local LinkedIn recruiting workspace on Browser Use's Jev Ultrafast: Jev screens titles, drives navigation, and assesses profile excerpts against your brief while a live browser stays visible; matches and evidence export as JSON.