Community · E2 · artifact verified

Judge player persuasion by each character's values

An npm library that lets game characters argue back: give a name, a persona and a goal, pass what the player typed, and Jev judges whether that character - with those values - was convinced; the same line can win a greedy merchant and offend an honest guard.

01 · Role in the system

What Jev does here

Honeytongue wraps Jev in a character-judgment API: the developer supplies a name, a persona and a goal; the library composes typed questions whose criteria are that character's values, sends the player's line, and reads back a validated answer - choice, score or noul - saying whether this character was persuaded and how strongly. Because the criteria travel with the character rather than the game, one line of dialogue lands differently per listener: argue profit with a greedy merchant and you win; say it to an honest guard and you offend. The client supports multiple models with retries, validates every answer's shape before returning it, ships with a browser play page and docs, and is published on npm as an alpha with CI running the test suite.

02 · Control boundary

Where Jev sits

Jev as the persuasion judge for typed game dialogue: character values become question criteria, the player's line is judged per listener, and validated typed answers - not free text - tell the game whether the NPC budged.

Code owns the loop, permissions, thresholds, validation, and side effects. Jev owns only the bounded judgments described above.

03 · Known limits

What this evidence does not prove

  • Alpha npm package; API surface may still change.
  • Persuasion outcomes are single-shot judgments; no multi-turn negotiation memory.
  • Both repo and site are new (0 stars at review time); no independent evaluation.

04 · Attribution

Public sources

This is a Community record: the project was published by a third-party community author.

  • tbrought ↗Community · github · public · checked 2026-09-29