User repository on GitHub
Jev judgments as TypeScript functions over arrays: classify, score, check, where, extract, pick, rank, verify. LLMs propose, Jev decides.
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Integrations & Adapters
Hunch JS
TypeScript build of hunch for Node 20 and up, mapping Jev's Choice, Score and Noul onto classify, score, check, where, extract, verify, rank and route over arrays and row objects.
- Stars
- 2
- Language
- TypeScript
- License
- MIT
- Published by
- Community
- Author
- steven-shoemaker
- Last verified
- Oct 2026
About this project
Judgment as a TypeScript function. Ask a question about a string or a whole array, and get an answer for every item that your code can branch on. Node 20 or newer. Also available for Python (pip install hunch-jev), with the same verbs. Using a coding agent? Give it the skill so it uses these verbs instead of writing prompt-and-parse code:
Quoted from the project's README. Read the full README.
Used inJev for Software & DevTools
- Languages
- TypeScript
Highlights
- One request per distinct item, sixteen in flight by default, duplicates asked once and several questions about an item batched into a single request.
- An array comes back in the same order and an object is treated as one row, so Jev sees every field unless you pass columns.
- detail: true adds the full distribution plus a shape, sure, split or unsure, so weak items can be routed to review.
- verify separates the source says otherwise from the source does not say, and text a claim puts in quotes must appear in the source word for word.
Watch out
Requires Node 20 or newer, and every verb needs a TypeSafe key. An object is one row and Jev sees every field unless you pass columns, so wide objects make wide prompts. Requests run sixteen at a time by default, so a large array fans out quickly against rate limits. As in the Python build, the README warns that predictions about behaviour sit near 0.3 to 0.4 when the item says nothing either way, so where is meant to be ranked with detail: true rather than filtered. The pandas accessors and DataFrame paths exist only in the Python package. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
Same verbs over the same three primitives, but the data model differs. One value returns one answer, an array returns an array in the same order, and an object is a row. The Python build additionally has pandas accessors and pairs over DataFrames. Concurrency is explicit here, sixteen requests in flight by default, and each verb also accepts a client if you do not want the default.
extract never writes a value. Code finds every candidate with the built-in finders, email, url, money, number, percent, phone and date, or your own RegExp or function, and Jev picks the candidate that answers the field from the words around it. Anything the text does not state comes back null rather than being invented.
Yes. ask takes a map of Classify, Rate and Check specs, sends every question about an item in a single request, and returns the answers as fields on each item, ready to join. Put context on an individual spec when only that question should see a policy, an ICP or a diff, so it does not colour the other answers.
Turn on detail and read the shape: sure when one option dominates, split when two are close, unsure when the evidence is flat. You can act on those directly, re-ask only the close rows with split: 'rematch', or let unsure come back as its own value. Cut-offs stay in your code, so check defaults to a threshold of 0.5 and route buckets take explicit probabilities.
Requests run in parallel, 16 at a time by default, and answers are cached. One request is made per distinct item: duplicate values are asked once, and several questions about the same item go in a single request.
Yes. Every verb also takes { client } if you would rather not use the default, and configure accepts apiKey, llm and cache settings, with llm available through anthropic, openai, azure, openrouter, ollama or a custom function.
pairs(a, b) lines up two arrays item by item so any verb can compare them. Use it to deduplicate records, match leads to accounts, or grade answers against references, then feed the result to a verb like score.
More Jev resources like this
8GitHub stars
Shadcn registry search block that shows lexical hits instantly then re-ranks the top candidates with Jev intent scoring in one request, with keyword fallback on failure.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
kylemclaren/jevsearchTypeScript
kylemclaren/jevsearchTypeScript0GitHub stars
n8n community node for TypeSafe's System One API, exposing yes or no, choice and score questions as workflow nodes, with an Evaluate Questions operation that batches several over one state in a single request.
Jev Integrations & Adapters#community#jevbest#catalog-import
69GitHub stars
Provider-neutral TypeScript runtime for System One decisions: write choice, noul, and score code once, then run it on TypeSafe Jev, local Reflex, any compatible /v1/systemone endpoint, or a custom provider.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
iamaamir/system-oneTypeScript
iamaamir/system-oneTypeScript38GitHub stars
File-first project notebook where many agents share context in .context/ files while Jev only judges the next step and records it as JSON.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
pedrocivita/tocketTypeScript
pedrocivita/tocketTypeScript25GitHub stars
Agent runtime with a terminal UI and a messaging gateway that keeps memory, self-built skills, session search, scheduling, and subagents across Telegram, Discord, Slack, WhatsApp, Signal, and CLI.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
wundercorp/lokiPython
wundercorp/lokiPython33GitHub stars
SillyTavern extension that sends chat text to Jev sensors and fires one rule action per turn such as nudging, rerolling, editing lists, or running scripts.
Jev Integrations & Adapters#community#jevbest#catalog-import
Community
mossyfield/ST-jevedJavaScript
mossyfield/ST-jevedJavaScriptFrom the Jev community
Posts from builders shipping with Jev right now.