Community · E2 · artifact verified
Watch novel characters take shape as you read
A reader for public-domain Japanese novels where every paragraph is judged by Jev and each character accumulates a trait radar for the page in front of you, plus a running ranking of traits for the story so far.
01 · Role in the system
What Jev does here
Reading Dazai's Run, Melos! in the built-in reader, you watch the tyrant king's radar start at scary, dictator and distrustful - and in the last pages forgiving and warm-hearted climb past them. Every paragraph goes to Jev as typed questions with probabilities and scores instead of prose judgments: which characters appear, what traits this paragraph shows for each, and how strongly. The page aggregates those judgments into a per-character radar for the current chapter and a cumulative trait ranking across the story so far, so character impressions emerge from calibrated paragraph-level readings rather than one summary verdict. The corpus is Aozora Bunko public-domain works, and the judge client is a typed TypeScript library.
02 · Control boundary
Where Jev sits
Jev as the paragraph-level judge in a reading companion: typed per-paragraph judgments over characters and traits, aggregated client-side into per-chapter radars and a cumulative story ranking - no prose, no summaries.
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
- Corpus is public-domain Aozora Bunko texts; Japanese only.
- Trait vocabularies and aggregation are the author's design; readings are not validated against human readers.
- Brand-new project (0 stars at review time).
04 · Attribution
Public sources
This is a Community record: the project was published by a third-party community author.
- nagoya0 ↗Community · github · public · checked 2026-09-29