Should I use threshold or at_least for uncertain email?

Use threshold for a single true-false cut defaulting to 0.5, where 0.51 and 0.99 both read true. Use at_least for a two-sided band that returns nil in the middle, and never pass both together.

Can I ask urgency, team, and severity in one call?

Yes. Batch several defined feels, decide, and score questions into one request and one state with Jev.measure, then read collapsed values or pattern-match without extra calls. Ad-hoc strings stay single-call only.

Why did decide return nil when a winner seemed obvious?

Decide returns the winning symbol, but confidence gating can return nil on weak calls. Score returns a float position for range matching, not a percentage, with stubs spreading neighbors to hit the mean.

How do I test without paying for live Jev calls?

Stub by definition name with probabilities, choices, and scores, restoring transport per thread. Record live answers to a tape matching state plus questions, then replay offline without network or stored keys.

How do I define a scored scale?

Jev.define with levels maps a symbol to a scale of 2 to 10 levels, ordered low to high, each with a description. score returns the fractional ordinal position between levels, not a percentage.

How many choices can one decision offer?

Up to 255 choices in a single decision. A symbol without extras is a yes-or-no check, choices make it a decide, and levels make it a score; a Symbol without a definition raises Jev::UndefinedDefinition.

What errors will I have to handle?

All errors inherit from Jev::Error. A missing key is Jev::ConfigurationError, HTTP 401 is Jev::AuthenticationError, 429 is Jev::RateLimitError, an unparseable body is Jev::InvalidResponseError, and a replay miss is Jev::ReplayError.

CommunityJev Integrations & Adapters

Jev Feels

Ruby gem that models Jev as field predicates: feels and decide bind to model attributes with shared definitions, thresholds, confidence gates, batching, and test tapes.

See highlights

Open github.com

Stars: 19
Language: Ruby
License: MIT
Published by: Community
Author: Qew7
Last verified: Oct 2026

Qew7/jev-feels

User repository on GitHub

View on GitHub

Semantic decisions as ordinary Ruby #feels?, #decide, #score, Rails validations and pattern matching powered by Jev

Forks: 1
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

email.feels?(:urgent) is a named question on a field, not a prompt in a service. The class is the scope, the first argument is the field. validatesfeeling accepts allownil:, allowblank:, if:, unless:, on:, message:, atleast:. If those skip the check, there is no HTTP call. A nil from atleast: is a validation failure. Jev::Model reads ticket.description (any method that returns a string). Bind-only uses the global Jev.define. Pass a string to define on the class: feels :description, :happy, "Sounds cheerful".

Quoted from the project's README. Read the full README.

Tagscommunity jevbest catalog-import awesomejev

Works forJev for Classification Jev for Ranking

Used inJev for Software & DevTools

Languages: Ruby

Highlights

Class-level feels, decide, and score bind a field method; validations support thresholds, allow flags, conditions, and messages.
Global plus class-scoped definitions cover yes-no, up-to-255 choice, and 2-10 level score vocabularies with symbol safety.
Measure batches defined questions into one state with collapsed hash output and direct Ruby pattern matching.
Match dispatches one decide with confidence fallback to otherwise; errors map HTTP and definition faults to classes.

Watch out

Score returns fractional position, not percent, batch uses only defined names, and replay matches request content rather than queue order with keys never stored. 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.

More Jev resources like this

loki

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.