typesafeai-review
Code reviewer where code runs checks and slices the diff, Jev answers narrow typed questions per hunk, and code composes the verdict.
Overview
typesafeai-review is a code reviewer built as ordinary software rather than an LLM agent. Code runs the checks and slices the diff into hunks, Jev answers narrow typed questions about each hunk, and code composes the verdict, score and findings. It ships as a standalone development dependency run with uv run review against a worktree, and a spec describes the design.
Most of the README covers the development environment: a devcontainer, credential volumes shared across projects and a single make check gate that runs controls, governance and tests. The specific questions asked per hunk are in the spec rather than the README. Maintenance commands include a single check gate covering controls, governance and tests, and credentials live in shared Docker volumes rather than bind-mounted host files.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Narrow typed questions about each diff hunk (e.g. does this hunk introduce a problem?)(inferred) | yes · no |
How it works
- Run code checks and slice the diff into hunks.
- Ask Jev narrow typed questions per hunk.
- Compose verdict, score and findings in code.
- Report the review.
Strengths
- Deterministic composition of model judgments.
- Per-hunk questions keep context small.
Caveats
- Questions and scoring are documented in the spec, not the README.
- Requires a TypeSafe key.
Getting started
uv run review --worktree <path>At a glance
- Type
- Project
- Status
- prototype
- Language
- Python
- Runtime
- CLI (uv)
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Using Typesafe.AI to generate diff reviews.
hellogumbo/awesome-jev
Frequently asked questions
Is it an LLM agent?
No, code orchestrates and Jev only answers narrow questions per hunk.
How do I run it?
uv run review --worktree <path>.
Where are the questions defined?
In docs/specs/typesafe-reviewer.md.
Related projects
Automated PR review: an LLM agent finds issues, a verifier refutes them, and Jev decides how to triage, route and label.
ProjectproductionTypeScriptRust code-review gate asking Jev small typed questions about functions and tests, failing CI only on rules measured 80% right.
CLIproductionRust- CodeOtter★ 3
Self-hosted PR review platform pairing an LLM narrative with a /v1/systemone decision engine for 6 rubric scores and 8 yes/no merge gates.
Servicebeta Coding-agent hook that asks Jev one yes/no per project rule on every edit or turn and makes the agent fix any violation.
HookbetaTypeScriptClaude Code plugin sending small judgments to Jev: rule checks that block edits, verbatim compaction, subagent sizing and routing hints.
PluginbetaPythonAdvisor for coding agents: git and Stop hooks flag test tampering, secrets and skipped hooks, with optional Jev review findings.
Hookbeta