Jev ecosystem projects — page 4

TypeSafe Mario is an experimental NES controller loop using emulator telemetry and RAM-derived game observations. With Python 3.13, a TypeSafe key and your local game setup, inspect state-demo before running the play command.

The repository collects separate latency-focused demos, including Sift email labeling and Turbo Rerank search ranking. Each app has its own setup: install dependencies in its directory, configure TYPESAFE_API_KEY and run its development command.

JevHarness lets a language model write task-specific features, questions and control logic, with optional improvement from rewards and complete traces. Explore the archived Pokémon run locally with the website scripts, or install the skill and describe a task to build a harness.

pg-jev filters, ranks and classifies PostgreSQL rows using plain-language SQL conditions. Batched Jev questions judge rows; SQL functions expose boolean, probability, choice and rubric-score results. Requires PostgreSQL 14–17, plpython3u, superuser access and a TypeSafe key.

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.

jev-voice-browser streams speech transcripts to a Node server controlling a visible Chromium window through Playwright. Jev selects intent and targets; code extracts literal text candidates and decides whether to act, wait or ask. Use Node 20+, a TypeSafe key and Chrome or Edge for the microphone control page.

LLM Debugger drives VS Code’s debugger, setting breakpoints, stepping and inspecting live variable values. The Jev strategy selects next steps and checks evidence sufficiency; a generation model supplies hypotheses, expressions and fixes. Configure model keys and use @debugger with a program or symptom; candidate programs are run to reproduce it.

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.

System One Adapter implements the typesafe_sdk evaluation interface using other providers’ LLM APIs. It enables comparisons with TypeSafe while preserving typed answers and adding retry, latency and normalization diagnostics. Configure a provider, model and probability or discrete answer mode; native structured outputs are optional.

AI Stock combines market data, news and AI analysis for stock research, screening and simulated trading. Optional JEV Decisions only returns buy/sell/hold, probabilities and confidence; research chat and stock selection still need an LLM. Configure models and data sources, select a stock and review the saved report’s evidence, dates and risks.

System One Connector exposes typed decisions to coding agents through its evaluate command and MCP integration. It connects to Jev or compatible self-hosted models such as CLM and Laya. Define Choice, Noul or Score questions; agent setup and batch evaluation are provided.

Unclutter hides nonessential elements in Chrome, Chromium and Firefox using saved rules for matching page templates. Jev classifies candidate elements into typed categories; extension code validates responses and controls selectors and reversible hiding. Choose TypeSafe or Vercel credentials and analyze manually by default. Saved rules continue working offline.

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.

This unofficial directory pairs public demos and repositories with author information, engagement metrics and source links. Its examples explore Jev's typed decisions in routing, classification, guardrails and games, including combinations with generative models.

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.

Define semantic expectations in perch.yaml, scoped to files or methods, and inspect the resulting code findings. The System One client asks typed questions about supplied state, defaults to jev-latest and calls TypeSafe's API.

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.

The tool extracts and indexes document content in Qdrant, plans an outline and researches each section against the local collection. Jev judges and reranks retrieved chunks and scores missing coverage; a separate LLM plans, queries and writes while Python controls retries.

This experimental CLI develops binary, multiclass, multilabel and scoring functions through repeated evaluation, labeling and definition proposals. Jev evaluates the data through TypeSafe, Vercel or Cloudflare; GEPA uses a separately configured reflection model.

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.

Jev Browser runs a task from a starting URL and returns the final page, screenshot and decision trace. Use the npm CLI, MCP server or navigate library with Node.js 22 and a Jev provider key.

The plugin uses Jev to recommend available tools, skills and task owners and to evaluate output against defined criteria. Install the native DSH bundle or import the iPolloWork package, configure a TypeSafe key, then check status before a live connection test.

Jev Codex Router chooses model capability and thinking effort for each Codex model call, including tool continuations. Use the documented macOS setup with Codex desktop, Node.js, Python and a TypeSafe key, then install the embedded stack and select its router.