Jev Projects Added This Week

LLM evaluation platform whose Instant Evals sends production traces to Jev for batched Noul, Score, and Choice judgments.

Open 4B decision model with prefill-only Choice, Noul, and Score inference over text or one image, plus native, vLLM, and SGLang runtimes.

Local 1B CPU decision model with Jev-compatible Noul responses; it returns P(yes) and explicitly rejects Choice and Score.

Local Qwen3.5 and Gemma 4 fine-tunes with a Jev-compatible API for typed Choice, Noul, and Score decisions in one forward pass.

Semantic code-search CLI for coding agents; Jev selects relevant repository branches, files, and source units for a natural-language behavior query.

Coding-agent workspace where Jev or a local Laya judge gates bounded decisions such as context retention, approvals, routing, drift, and inter-agent sharing.

Local-first coding workspace where Jev or Laya selects an eligible agent/model route and the host validates every choice before execution.

Reasoning Jev-style 9B classifier with a diffusion drafter and compatible API for typed Noul, Choice, and Score decisions.

Multimodal Qwen3.5 decision models that answer multiple typed Choice, Score, and Noul questions with probabilities in one forward pass.

Apple-silicon computer-use toolkit that runs local typed-decision models to choose form-fill actions from macOS accessibility state.

Cross-platform chat assistant where Jev chooses a bounded reply strategy from reviewed conversation context before the user sends any draft.

Paper-trading race where Jev chooses each bee's valid market move and conviction while deterministic code enforces risk limits and execution.

Local multimodal decision models that answer Jev-compatible typed questions over records, text, and up to two images, with explicit abstention.

Installable agent skill that teaches Jev typed-decision design, validates live API shapes, and indexes reusable patterns from community projects.

Project-memory tooling for coding agents; Jev decides which turns to save, their kind, contradictions, recall relevance, and suspected memory poisoning.

Pokémon Red harness where Jev chooses every menu, navigation, team, and battle action from legal options exposed by the emulator.

Specialized CUDA inference backend for Open-Jev-27B with a drop-in Jev-compatible server, prefix-tree reuse, and checked benchmark artifacts.

Claude Code skill and CLI that delegates bulk ranking, classification, filtering, and semantic-search judgments to Jev.

Serving runtime for open Jev-style text and vision checkpoints using vLLM on Linux or MLX/MPS on Apple Silicon behind /v1/systemone.

Framework and fine-tuned embedding models that turn cosine similarities into Jev-style Choice, Score, and Noul probabilities.

Frozen open-model medical decision stack that scores each exam option, then uses confidence and conformal gates to auto-release or escalate answers.

Chinese 11-chapter hands-on course with runnable notebooks for Jev primitives, recipes, evaluation, applications, agent integration, and local-model training.

Open 4B English/Korean decision model with Jev-compatible Choice, Score, and Noul probabilities plus BF16 and GGUF weights.

Scheduled research-paper radar that asks Jev whether each paper matches configured interests and exclusions, then publishes ranked web and feed output.

PHP 8.2+ client for Jev with typed Choice, Score, and Noul request builders, result classes, retries, and base-URL overrides.

PowerShell module that submits typed Noul, Choice, and Score questions to Jev and returns pipeline-friendly structured answers.

Local 0.8B Jev-style model and compatible server for Choice, Score, and Noul decisions, with agent skills and a Claude Code permission guard.

Reproducible paper-trading arena that gives Eikos and Jev identical market states and typed questions, then compares equity and probability forecasts.

Open multimodal decision model that returns candidate probabilities for text and screenshots through a Jev-style Choice, Score, and Noul endpoint.

Browser playground for editable Jev Choice, Score, and Noul questions through TypeSafe, OpenRouter, or Vercel AI Gateway providers.

Reproduction kit for a 38-benchmark typed-decision index with frozen public-source suites, resumable runners, official scorers, and Jev-compatible HTTP engines.

Open 2B and 4B multilingual decision models focused on Korean, serving Jev-compatible Choice, Score, and Noul probabilities through Kev.

Browser driving game that captions dropped road images locally, then uses Jev Choice answers to drive the car while code owns physics and timing.

Evidence-backed directory of Jev robustness, calibration, invariance, injection, abstention, and multilingual studies with machine-readable entries.

Open multimodal System One model in four sizes that answers typed questions over screenshots, images, video, or text in one forward pass.

Self-hosted Jev-compatible server for Laya and Kev engines with Docker images, official-SDK compatibility, and runnable decision games.

.NET 10 client for hosted Jev and compatible local servers, with typed Choice, Score, and Noul answers plus Microsoft AI integrations.

Oh-my-pi extension that asks Jev batched Noul relevance questions about tool-output and earlier assistant-text chunks before main-model context, with opt-in filtering and recoverable originals. No repository-level license file was found during review.

Local decision stack that fuses four readings from two frozen open models, then uses agreement and calibration gates to answer or abstain.

Provider-independent .NET decision library whose Jev provider maps typed Choice, Score, and Noul questions to the System One API.

Local logprob decision layer and Claude Code permission gate that scores command risks, routes model tiers, and judges retries or stuck runs.

Research-agent integration where Jev locates relevant source regions with Choice and Noul, then code returns verbatim passages for verification.

Drop-in proxy and Pydantic AI client that records Jev probabilities and alerts on model-version changes, distribution drift, threshold crowding, and estimated accuracy loss.

Shared emoji canvas where Jev chooses which emoji and square to paint next and judges whether to finish the visitor's stroke.

Choice-only decision exchange that routes bounded business decisions to TypeSafe Jev or compatible providers and maps validated results to non-authorizing action intents.

Bilingual Jev tracker and interactive playground whose templates submit typed Choice, Score, and Noul questions through a Cloudflare proxy and display structured results. No repository-level license file was found during review.