Écosystème Jev
Un petit répertoire de SDK officiels, clients communautaires, intégrations, outils d’agents et Playgrounds vérifiés.
Officiel
TypeSafe JavaScript SDK
Official TypeScript/JavaScript client for System One requests, including Choice, Score, and Noul builders with inferred answer types.
Vérifié le: 2026-09-21
TypeSafe Python SDK
Official Python client for TypeSafe's System One API, useful when Jev decisions live in Python services or data pipelines.
Vérifié le: 2026-09-21
System One Adapter (Python)
Official adapter that exposes a TypeSafeClient-compatible interface backed by chat-model providers for side-by-side workflow comparisons.
Vérifié le: 2026-09-21
TypeSafe Agent Skills
Official agent skill material for using System One primitives and evaluation patterns from coding agents such as Codex and Claude Code.
Vérifié le: 2026-09-21
SDK
advocaat
A compact community client that wraps TypeSafe/Jev decisions behind a type-oriented developer API.
Vérifié le: 2026-09-21
typesafe-ai (Rust)
Community Rust client for TypeSafe AI with typed request/response handling for System One workloads.
Vérifié le: 2026-09-21
typesafe-sdk-go
Community Go SDK for the TypeSafe API, aimed at typed questions and probability-bearing answers in Go services.
Vérifié le: 2026-09-21
typesafe-sdk-swift
Community Swift SDK for TypeSafe AI, providing a native path for Apple-platform applications to call System One.
Vérifié le: 2026-09-21
Intégrations
Vercel Eve
Vercel's open agent framework; its experimental evaluation path includes Jev as a decision/evaluation model option.
Vérifié le: 2026-09-21
Vercel AI CLI
A Vercel Labs terminal tool whose evaluation workflow can use Jev as the evaluator.
Vérifié le: 2026-09-21
zod-jev
Combines local Zod shape validation with Jev-powered semantic judgments so application code can validate structure and meaning separately.
Vérifié le: 2026-09-21
zio-typesafe-ai
Scala 3 / ZIO integration for the System One API with typed multi-question calls.
Vérifié le: 2026-09-21
Playgrounds
TypeSafeAI Community Playground
A large community playground exploring Jev patterns, workflows, routers, browser-agent experiments, and evaluation-oriented examples.
Vérifié le: 2026-09-21
Jev Playground
A community playground focused on trying Jev primitives and inspecting structured outputs from a browser-oriented interface.
Vérifié le: 2026-09-21
Agents / outils
Jev Ultrafast
Browser-use experiment where Jev selects browser operations and targets from a constrained action space while text generation is handled separately.
Vérifié le: 2026-09-21
Foreman
Agent-supervision tooling that uses Jev-style decisions to help keep coding-agent workflows aligned with their current task.
Vérifié le: 2026-09-21
jev-review
A community staged code-review workflow and local dashboard that uses Jev for bounded review decisions.
Vérifié le: 2026-09-21
jev-mcp
An MCP server that exposes Jev judgments to coding-agent and MCP-client workflows.
Vérifié le: 2026-09-21
jevwire
A decision layer for agents that combines an MCP surface, an embeddable library, and escalation-oriented integrations.
Vérifié le: 2026-09-21
Agent Chaperone
Screens agent tool calls and tool results around the execution boundary, providing a useful example of semantic checks surrounding deterministic controls.
Vérifié le: 2026-09-21
Canny
Open-source agent-completion tooling that focuses on evidence before accepting claims that a coding task is finished.
Vérifié le: 2026-09-21
jev-router
Open-source LLM-routing project that uses Jev to select a model from a bounded catalog on top of LiteLLM.
Vérifié le: 2026-09-21
Open source / recherche
Awesome Jev
A community-maintained, searchable directory of Jev projects that JevHub uses only as a discovery source before checking original projects.
Vérifié le: 2026-09-21
SemIf
Independent research project exploring Jev-like semantic if-statements with open models; useful as evidence that the decision-model pattern extends beyond one provider.
Vérifié le: 2026-09-21
Comment le répertoire est maintenu
- Chaque entrée doit avoir un lien public vérifiable.
- Les descriptions sont écrites par JevHub, sans copier une autre liste.
- V0.1 ne classe pas les projets par étoiles ou popularité.
- La sélection est manuelle, sans importation automatique.
Découvrez comment Jev prend des décisions ou commencez avec les huit modèles. Qu’est-ce que Jev ?、Modèles。