devagrawal09/jev-review

A staged code-review workflow and local dashboard built with TypeSafe Jev.

What it solves

Jev Review is a code-review workflow that automates the process of scanning Git diffs or entire codebases for issues. It helps developers identify potential problems in correctness, security, reliability, compatibility, and test coverage without requiring a full static analysis suite.

How it works

The tool uses TypeSafe Jev to perform "bounded judgments" through a series of focused model calls. It follows a structured risk matrix to classify mechanisms, select evidence, and score severity. The system can either review specific changes in a Git diff or scan all non-ignored source files in a codebase. It then presents these findings in a local, private dashboard.

Who it’s for

Developers who want a structured, AI-driven approach to code review and codebase scanning that operates locally and focuses on structured signals rather than generic AI summaries.

Highlights

  • Dual Review Modes: Supports both targeted reviews of current Git diffs and comprehensive scans of entire codebases.
  • Structured Judgments: Uses a risk matrix and explicit decision boundaries to score impact and classify issues.
  • Detailed Context: Incorporates changed or related tests as context when evaluating test gaps.
  • Local Dashboard: Provides a quiet, local-only HTTP server to view large reports in collapsible sections.
  • Strict Architecture: Built with a layered architecture that prevents circular dependencies and upward imports.

Related

  • Project
    chalk-lab/Mooncake.jl

    A high-performance automatic differentiation package for Julia that supports mutation and leverages optimized intermediate representations for efficient derivative calculations.

  • Project
    btsouth/omakade

    A unified Linux game library manager that aggregates games from Steam, Lutris, Heroic, and emulators into a single, cover-focused interface.

  • Project
    messkan/prompt-cache

    A lightweight, self-hosted semantic cache for GenAI workloads that reduces LLM costs and latency by serving cached responses for similar prompts.

  • Project
    CederGroupHub/chgnet

    A pretrained universal neural network potential for charge-informed atomistic modeling that predicts energy, forces, and stress with near-DFT accuracy.