User repository on GitHub
Decisions as code on System One models: typed questions, thresholds with hysteresis, nested decisions, journal, calibration
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Repos & SDKs
Huncho
Dependency-free TypeScript SDK that turns Jev Noul, Choice and Score answers into named decisions: enter/exit thresholds with hysteresis, nested decision trees settled in one call, a JSONL journal, threshold replay without inference and Brier calibration.
- Stars
- 1
- Language
- TypeScript
- License
- MIT
- Published by
- Community
- Last verified
- Oct 2026
About this project
A hunch is a probability with a policy attached. huncho makes that a first-class object in TypeScript: ask typed questions of a decision model, apply thresholds that do not flap, nest decisions, journal every one, replay a policy change without inference, and calibrate against what actually happened.
Quoted from the project's README. Read the full README.
- Languages
- TypeScript
- Platforms
- OpenRouterVercel AI Gateway
- Capabilities
- Open sourceRequires API key
Highlights
- Published on npm at 0.6.0 with zero runtime dependencies; the public API has been frozen since 0.1.0 per the changelog
- Hysteresis example: a decision enters page at 0.8 and holds there until urgency drops below 0.6, tracked per key
- Nested decisions settle in one model call; answer types follow from the questions, so using a non-existent option is a compile error
- The journal records every decision, and replay applies a new threshold to recorded answers with no inference
- Providers out of the box: TypeSafe direct, OpenRouter and Vercel AI Gateway; a PyPI port against the same fixtures is planned
Watch out
MIT license. Version 0.6.0 predates 1.0 and the repo had 1 star at verification; the PyPI port is announced as next but not yet shipped.
Questions about this entry
Short answers drawn from the linked project or page.
A probability with a policy attached: typed questions produce answers, thresholds map them to named outcomes, and each decision records whether it was reached via enter, hold or else.
For the default jev() provider yes, read from TYPESAFE_API_KEY. The OpenRouter and Vercel AI Gateway variants accept their own keys, and the model seam allows custom providers.
With enter and exit thresholds: a rule such as { enter: 0.8, exit: 0.6 } enters its outcome at 0.8 and stays there until the probability drops below 0.6 for that key, instead of bouncing on every crossing.
Yes. Decision models sit behind a seam, with TypeSafe Jev direct, Jev through OpenRouter, Jev through Vercel AI Gateway, and a factory for anything else; nothing above the model seam knows which vendor answered.
No. It is published as zero-runtime-dependency TypeScript, and the public API has been frozen since 0.1.0, with the changelog recording every change.
Yes. Every decision is journaled, so you can replay a threshold or policy change without inference, then calibrate the result against what actually happened.
A PyPI huncho port is in progress, built from the same fixtures now that the TypeScript API is frozen; each port passes the same wire and policy files, so journals are interchangeable.
Related terms
Glossary definitions related to this entry.
More Jev resources like this
2GitHub stars
Ruby gem that makes System One measurement a language primitive: psi prepares a state, judge/choose/score return calibrated distributions, and ? collapses them for control flow. Ships TypeSafe, protocol (/v1/systemone) and CUA providers behind one interface with its own spec suite.
Jev Repos & SDKs#community#ruby-gem#sdk
Community
innocentdiaz/s1_rubyRuby
innocentdiaz/s1_rubyRuby1GitHub stars
Python library that records each Jev-class decision as the full distribution under a sha256 key of state, question, provider and model, then applies a separate policy. Thresholds, review bands, routing and budgets can change across the recorded history without new model calls.
Jev Repos & SDKs#community#calibration#policy
Community
YYTbit/metajevPython
YYTbit/metajevPython12.4kGitHub stars
Open-source AI email assistant (Next.js, Postgres, 12k+ stars) with a decision-model layer that can run on openrouter:typesafe/jev-1.13: Jev-backed questions cover rule selection, cold-email detection, thread status, recurring patterns, unsubscribe pages, and sender categorization, each with evals.
Jev Integrations & Adapters#community#email#gmail
Community
elie222/inbox-zeroTypeScript
elie222/inbox-zeroTypeScript4.9kGitHub stars
LangWatch Instant Evals ships a TypeSafe Jev classifier that batches Noul, Score, and Choice judgments on production traces in one POST per text.
Jev Benchmarks & Evaluations#community#typescript#evaluation
Community
langwatch/langwatchTypeScript
langwatch/langwatchTypeScript882GitHub stars
A Node and TypeScript package that runs Convai's Laya checkpoint through ONNX Runtime, with the same system_one shape as the Python reference.
Jev Repos & SDKs#community#typescript#open-models
Community
receptron/layaTypeScript
receptron/layaTypeScript4GitHub stars
A pre-release Python framework for predictions on mixed text and tabular data: Jev Noul and Score questions turn text columns into numeric features, then a tabular foundation model such as TabPFN learns from labeled rows in context. Offers hosted and fully local backends.
Jev Repos & SDKs#community#tabular#predictions
Community
edamame-labs/tab-jevPython
edamame-labs/tab-jevPythonFrom the Jev guides
Original write-ups on what this entry is for.
Closed weights have not stopped the ecosystem from rebuilding the shape in the open. API-compatible servers, retrained heads on small models, and inference-time tricks that skip generation entirely.
Jev Repos & SDKs#open-models#research#local
Read article
From the Jev community
Posts from builders shipping with Jev right now.