537 repositories with documented relationships and source evidence.
Laya uses a non-autoregressive model to answer choice, score and noul questions, with a router selecting checkpoints. Its self-hosted server exposes POST /v1/systemone for existing Jev clients; the underlying model is Laya rather than hosted Jev.
Kev provides pretrained decision models and fine-tuning for yes/no, multiple-choice and rating questions over shared text. Its API follows TypeSafe System One, allowing the TypeSafe Python SDK to target a local Kev server.
SemIf handles small semantic decisions such as routing, retries and evidence checks by reading option probabilities without generating an answer sentence. It reproduces Jev's interface pattern with open models, not Jev's undisclosed model or training, and has no TypeSafe affiliation.
The model answers choice, Boolean and ordered-score questions from states and candidates without decoding answer tokens. NanoJev is a separate local model inspired by Jev; recorded comparisons include Jev and untuned Qwen on the same game interfaces.
Give jevos text or structured state and a yes/no question; it returns the probability of yes. The local server uses Jev's wire format for Noul requests, allowing existing clients to address a separate self-hosted model.
RAG-Fusion generates query variants and merges retrieval rankings with Reciprocal Rank Fusion to find material a single query may miss. Jev is an optional reranker in the evaluation pipeline; its experiments require JEV_API_KEY. The original demo remains vector-only. Compare retrieval strategies on NFCorpus with the evaluation harness; the hybrid pipeline combines BM25, vector search, fusion and reranking.
decider returns Choice, Score and Noul probabilities from a state and typed questions in one forward pass. This independent Jev-style model family uses Qwen bases; its training data was not distilled from Jev. Call the Python Decider interface or point TypeSafe SDKs at its compatible local HTTP endpoint.
AnyJev reads probabilities from open models to answer choice, yes/no and scoring questions without generating prose. It is an independent Jev-style approach; its L2 heads leave the base model’s weights unchanged. Use label-free L0 readouts or fit an L2 head for a particular question, with transformers or vLLM.
TypeLLM adds JSON Schema-constrained outputs to existing autoregressive models without changing their weights or architecture. Inspired by Jev, it retains free text and optional reasoning and supports dependencies between output fields. Serve a compatible model with SGLang, then submit context and typed questions through TypeLLMClient.
Rizzo Flow runs open decision models locally, returning choices, probabilities, rubric scores and numbers through a Jev-compatible HTTP interface. It reproduces Jev’s interface pattern, not its proprietary architecture or training; the project is independent. Serve a selected model with llama.cpp and point TypeSafe-compatible application code at the local endpoint.
Von scores independent questions in one forward pass for classification, yes/no checks and ordered ratings. It is a separate System One model with a TypeSafe-compatible API, running in-process or through HTTP. Use the Python or TypeScript SDK, or expose the local model with von serve.
LLM2Jev scores candidates from local language and vision models using prefill logits without token-by-token generation. It provides an independent Jev-style implementation with a compatible System One HTTP endpoint. Use Python or HTTP with SGLang, Transformers or MLX, including supported text-and-image inputs.
Valen scores supplied candidates against text, images or video and returns decision probabilities without generating answer tokens. This Jev-inspired implementation uses a Qwen3.5 backbone and a shared decision head. Run inference or evaluation with both the Valen checkpoint and its Qwen3.5-2B base model.
The project pairs Qwen3.6-35B-A3B on SGLang with a separate Python API process for typed decision serving. POST /v1/systemone supports Noul, Choice and Score; jev-latest is an alias for the configured Qwen model, not a TypeSafe request.
A candidate-scoring head returns distributions for gates, routes and scores without decoding prose or JSON answers. The README compares a benchmark-trained AgentJev specialist with Jev 1.13.0 evaluated zero-shot; these use different training conditions.
The platform supports developing and evaluating SRE agents against live system environments containing operational failure scenarios. Jev support is disabled by default; when enabled for Codex, it reviews diagnostic tests and submissions from evidence sent to TypeSafe.
This learning project answers choice, yes/no and ordered-score questions about one image by directly scoring candidate outputs. It is an independent community exploration of Jev-like inference, without reproducing TypeSafe's proprietary architecture, training, calibration or serving system.
The repository offers open decision models for Jev-style Choice, Score and Noul workflows, including separate Verdict and Verdict 2.0 models. Serve webgpu-demo with Python’s HTTP server and open the documented browser playground; model artifacts and evaluation receipts are linked separately.
JevBench v1 - a benchmark for Jev-class typed decision models: smart, cheap, fast, reliable, open.
A small open decision model: state + typed questions -> calibrated probabilities. A Jev / System One re-creation on Qwen3.5.
Leaderboard that puts Jev and 14 LLMs through 108 four-option multiple-choice questions about the Convex backend platform, comparing accuracy, latency, and cost, with a per-answer explorer showing Jev probabilities.
JevK5: open-weight alternative to TypeSafe Jev. Typed decisions with probabilities in one forward pass; Apache-2.0 weights and code.
Open Jev implementation with custom finetuning
Agentic RL 中文零基础教程（33 章）：从概念到 GRPO 实战，含 TRL 最小可跑示例；26–33 章附一套可运行的三方判别模型实证工程（encoder vs LLM-LoRA vs 规则基线）。第 25 章讲清 Jev / TypeSafe System One 与 RL 的能力边界 | Chinese Agentic RL tutorial (33 chapters) + a reproducible discriminative-model benchmark