These are the examples that live in a public GitHub repository: tools, SDKs, MCP servers, cookbooks and projects built on Jev and other decision models. A decision model returns a typed choice or score with a confidence value, so code can act on it without parsing text. Repositories here range from small client libraries, such as AbdelStark/s1-rs for Rust, to full agents and apps, such as agent-labs-dev/fastbrowse and clownware/bouncer. Many are one-person projects. Before you depend on one, open the repository and check the licence, the date of the last commit, the open issues and how the README describes setup. Star counts are only a rough guide. Each entry here links to its source, so you can read the code before you run it.
430 repositories
Showing 1 to 48 of 430
Browser Use's fast browser agent where Jev picks the operation and the DOM element in a single request; a small LLM is only invoked to write text when typing is required. Demoed on a Google Flights search.
19k starsFlight search 7.1 s
Attempt to run a Jev-style System One decision model locally on a single RTX 3090. Companion site at openjev.com. The repository was renamed to TheoLeeCJ/SemIf on GitHub, and the old openjev URL now redirects to it.
TheoLeeCJOpenopenjev on github.com4.1k starsJSON vs direct 5.21x slowerPeak throughput 20.03 dec/s
Makes one AI trade decision every Monad block on the Kuru MON-USDC market using Jev. The public deployment runs in dry-run mode with a mock model.
jarrodwattsOpenjev-trader on github.com2.2k starsLoop, mock model p50 100 msDecision window ~300 ms/block
Official agent skill for Claude Code, Codex and other agent environments; installs via claude plugin marketplace add typesafe-ai/skills or npx skills add typesafe-ai/skills. MIT licensed.
Open reimplementation that trains a small model to choose among a changing list of text options, emitting one probability per option in a single forward pass. Ships Doom, chess and Wikispeedia demos.
Independent community collection of use cases, patterns, prompts and starter code for Jev, explicitly not official. It also links sibling collections by the same author and two related projects that report their own cost figures.
MCP server that gives Claude Code, Codex, Amp, OpenCode and other agents ten Jev judgment tools, each returning typed probabilities they can branch on. The author calls it early software.
320 starsPer call 150-500 ms
Declares typed input/output Signatures for LLM modules and optimizes the underlying prompts and weights against a metric. Its Signature abstraction is the closest widely-used open equivalent of Jev's typed-question interface. Star count is GitHub's rounded display figure.
Cua's own family of small System 1 models for computer-use decisions, separate from TypeSafe's Jev. CUA-S1-FORMS is the first release, with weights and a dataset on Hugging Face.
Decision model from Convai Innovations that fully fine-tunes an encoder and adds a typed-decision head: a 421M English checkpoint on ModernBERT-large and a 322M multilingual one on mmBERT-base, with a router picking between them by detected language. Returns choice, score and noul answers, and reports expected calibration error of 0.081 after temperature fitting, down from 0.466 as shipped.
Guarantees valid structured output during generation by constraining decoding to a grammar or schema. The mechanism HN commenters repeatedly cited as already covering Jev's cannot-produce-a-type-error guarantee on ordinary LLMs. Star count is GitHub's rounded display figure.
Pydantic-based library that extracts typed, validated structured data from LLMs across providers with automatic retries. The established way teams get Jev-style typed answers today, at LLM latency and cost. Star count is GitHub's rounded display figure.
Claude Code plugin that replaces the compaction summary with Jev decisions. Every tool call and result is scored in one request against the whole conversation, stale ones are dropped or truncated, and everything kept stays verbatim. Also usable as an npm library.
A family of small, self-hosted decision models built on Qwen3.5 at 0.8B, 4B and 9B, with training code and an API that matches TypeSafe's System One so TypeSafe's own Python SDK can point at a local server.
An independent MLX port of Laya for Apple Silicon, 13.4ms median for a short English decision and 7.4ms on the multilingual checkpoint, with a Snake demo that plays at up to 75 moves per second under 1GB of memory.
mizorewwwOpenLaya-MLX on github.com6.1k starsShort decision 13.4 ms P50Snake demo 75.4 moves/s
Compact encoder model that takes the label set as input at inference time and extracts arbitrary entity types zero-shot on CPU. Same classifier-that-takes-its-categories-as-input shape, for extraction rather than decisions. Star count is GitHub's rounded display figure.
Fine-tunes Sentence Transformer embeddings plus a lightweight head for high-accuracy classification from a handful of labeled examples, with no prompting. The standard cheap alternative to an LLM classification call when the label set is fixed. Star count is GitHub's rounded display figure.
NanoJev is a 0.6B open-source replica of Jev that takes states and questions and returns complete probability distributions with zero output-token decoding. Ships with model weights, dataset and an end-to-end training pipeline, plus a side-by-side demo running Jev, NanoJev and untuned Qwen through Snake and a 50x50 maze.
Modernized BERT encoder used as the backbone for fast fine-tuned classifiers and rerankers. The non-generative baseline Jev's cost and latency claims are usually measured against. Star count is GitHub's rounded display figure.
Rust desktop automation that reads apps through the OS accessibility tree instead of pixels. Its jev-desktop skill takes a plain-language goal and has Jev pick one operation and one target element per turn, so the 150-element tree never enters the agent's context.
1.6k starsToken reduction 78-96%
Core ML ports of all three Laya checkpoints for Apple Silicon, reaching 4.98ms P50 on the Neural Engine for a short multilingual decision, with a Snake demo sustaining about 50 decisions per second.
mizorewwwOpenLaya-CoreML on github.com1.4k starsShort decision 4.98 ms P50Snake demo 49-50 dec/s
macOS computer-use loop that OCRs the screen, classifies the next action with Jev, then clicks. Author reports roughly $0.0002 per step.
Aaron Levin (awlevin)Opentypesafe-computer-use on github.com935 starsCost/decision $0.0002, 155xLatency 0.13-0.38s
Vercel Labs CLI for generating from the terminal. Its evaluate command takes a state plus typed questions and defaults its evaluation model to typesafe-ai/jev on the AI Gateway, with -m jev as the shorthand and an --input flag for text or JSON state.
Agent skills for Hermes, Claude Code and Codex that hand Jev the routing decisions: which model tier to use, which search results to open, which memory passages to read and which hide injected instructions. Ships a dashboard with an on, shadow and off switch and publishes measured latencies per skill.
NeurIPS 2024 masked discrete diffusion language model that produces outputs in parallel rather than token by token. Cited in Jev coverage as the open non-autoregressive research line closest to Jev's parallel sampler.
Python library from Nokia researchers that reads typed Choice, Noul and Score answers with probabilities from an open LLM's logits or hidden states, without fine-tuning. Its zero-label level reduces option-order bias, and a head fit on 100 to 300 labels can run behind vLLM.
A macOS menu bar companion with two modes: type a task and Jev picks what to click from a local OmniParser running on CoreML, looping for up to 12 actions, or talk to Gemini Live, which takes one desktop action per turn.
Milind S (@milindlabs)Openmilind-soni/tiptour-macos on github.com662 starsAction limit 12 actions
Apache 2.0 decision model on a 395M ModernBERT encoder that serves the same /v1/systemone API as Jev, installed as von-sdk from pip or npm. Version 1.2 scores each option on its own, so reordering options no longer changes answers. Its own table puts Von 1.1 at 72.0% on the 49-task jabr v2 suite against 96.6% for Jev.
Staged code-review workflow with a local dashboard, where Jev judges code changes or a whole codebase.
Places Jev as a fast supervisor above slower Codex coding agents. Codex workers do the engineering while Foreman independently assesses whether the work is complete, requirements are met, tests suffice, or human input is needed.
A Minecraft agent that pairs an Astra planner with Jev as the real-time controller, with native run recording, tested routes and run verification.
536 starsFastest run 8m 43s, -40%Decisions used 131 Jev calls
Curated list of official resources and community projects for TypeSafe, System One models and Jev, with a GitHub Pages site. Sections cover docs, SDKs, agent tooling, browser agents, games/robotics, evaluations and showcases.
Deploys DiffusionGemma-Jev (djev) as a Jev API-compatible endpoint on Cloud Run with one RTX PRO 6000 GPU, scaling to zero when idle. The Gemma team's video demo reports 35 to 60 ms single-step latency at roughly $3 an hour.
Google Gemma (@googlegemma)496 starsOpendjev-run: DiffusionGemma-Jev on Google Cloud Run on github.com
A browser skill for Codex where Jev handles navigation, clicks, toggles and scrolling while Codex stays in charge of text input, visual judgment and the final check.
447 starsSpeed 5-10x faster
Web search in plain language where Jev chooses sources, time ranges and search terms, then scores each result for relevance, with retrieval through Search1API. Returns links and snippets with visible relevance scores and editable filters, and writes no answers.
A TypeScript classifier that sends each PDF page's text to Jev as a choice over 261 IRS forms and 7 page kinds, with a second call only for five corporate forms and their schedules. Its eval reports 0 wrong pages on 314 filled TaxCalcBench forms at about $0.001 per page, 34x cheaper and 6x faster than the Sonnet pipeline it replaced.
Nakshatra Saxena (@nedwize)Openkyotofin/tax-doc-classifier on github.com419 starsCost per page $0.001vs prior setup 34x cheaper
Agent that plays Super Mario Bros. by feeding structured emulator state to Jev and acting on the returned controller action.
Standalone Android agent for Mobilerun where Jev makes every action decision, with a React studio UI and an Uber booking demo.
droidrunOpenmobile-jev on github.com377 starsUber demo ~21s, 9 actions
Automatic per-turn model routing for Claude Code and OpenAI Codex. Jev sends simple work to the fast model tier and difficult work to the strong tier while preserving each CLI's native tools, sessions, permissions and authentication.
Node.js and TypeScript package that runs the Laya decision model through ONNX Runtime, so JavaScript projects can call it without Python.
receptronOpen@receptron/laya on github.com374 stars3-question call ~140 msModel size ~1.7 GB
Open reproduction of the System One model class, fine-tuned from Qwen3.5-2B-Base, returning a probability distribution per typed question in a single forward pass. Serves TypeSafe's POST /v1/systemone wire format so TypeSafe SDKs work unchanged by repointing TYPESAFE_BASE_URL.
Curated list of Jev / TypeSafe System One applications, libraries and resources.
PostgreSQL extension that lets you filter, rank and classify rows with plain-language conditions, such as WHERE jev(tickets, 'the customer is angry'). Adds jev_prob, jev_choice and jev_score functions that compose with joins, GROUP BY and ORDER BY, with no index or embeddings. Published on PGXN.
realZachiOpenpg-jev on github.com329 stars2k rows, cold ~3.5s, $0.0122k rows, cached ~50ms
Prefill-only server that implements the TypeSafe Jev HTTP API on open models, running Qwen3.6-35B-A3B on SGLang. Deploys to Modal on a B200 with radix caching and breakable prefill CUDA graphs, scaling to zero after five idle minutes.
MCP server for TypeSafe shipped as a single static Go binary with no Node or Python runtime. Sends usage guidance to the client so the agent writes better questions.
Official drop-in replacement for TypeSafeClient that sends the same typed questions to ordinary LLM APIs, so developers can compare Jev against an LLM on cost, speed and answers.
A CLI from Sutro for building AI functions with Jev: it finds uncertain examples, has you label them, and uses GEPA to improve the function.
Nothing matches those filters
Clear a filter to widen the search, or send us the thing you were looking for.
