convez.ai
SkillEscrita & ConteúdoAvançado
Rayne-idk

Lorebook Builder

Rayne-idk/lorebook-builder

Extrai e estrutura lore de wikis em lorebooks prontos para uso

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill do Claude Code que pesquisa universos ficcionais através de wikis e gera lorebooks completos em formato SillyTavern/Chub com personagens, locais, facções, raças e histórico. Automatiza a coleta e síntese de informações de múltiplas fontes web para criar bases de conhecimento estruturadas.

Instalação

$npx convez-ai Rayne-idk/lorebook-builder

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
1
Forks
0
Acompanhando
0
Saúde
28%
Último commit
há 5 meses
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Construir lorebooks para universos ficcionais a partir de wikis existentes
  • Documentar worldbuilding para roleplay em SillyTavern e plataformas similares
  • Sintetizar informações espalhadas em múltiplas páginas web em formato estruturado

Veredito

Ferramenta especializada e bem-engenharia para curação de lore, mas com custo tokenômico alto — ideal para projetos específicos, não para uso casual. Requer integração Firecrawl e planejamento de orçamento de tokens antes de executar.

README

A tradução para o português deste README ainda está na fila.

Lorebook Builder

A Claude Code skill that researches a fictional universe through its wikis and produces a comprehensive SillyTavern / Chub-format lorebook (World Info) — characters, locations, factions, races, history, and lore, all written from real wiki sources.

Compatibility: built and tested against Claude Code. The skill is plain markdown (instructions + a standalone Python script), so it should be portable to other agent frameworks with a similar skills/tools mechanism — but that has not been tried.

⚠️ Token usage warning

Running /lorebook consumes a very large number of tokens. This is inherent to the task, not a bug: the model has to read through dozens to hundreds of wiki pages and condense an entire fictional universe into prose, then write 20–400+ dense lorebook entries from that research. A single run is comparable to summarizing a small book, and large/sprawling franchises (see the size bands in WORKFLOW_GENERAL.md) cost proportionally more and take a long time to finish. Keep this in mind before pointing it at a huge, long-running series.

With the current version of the skill, a full run (example: example-results/haiku.solo-leveling.lorebook.json) used about 25% of a Claude Code Pro subscription's 5-hour usage budget, using Haiku as the model. Using Sonnet resulted in 142% usage. You can compare the lorebooks they produced in example-results/.

Requirements

  • Claude Code
  • Python 3 (standard library only — used by the build/validate script)
  • Firecrawl skills configured (preferred research path); falls back to built-in WebSearch/WebFetch if unavailable, but is slower and less thorough. Firecrawl's free trial covers a generous number of scrapes, which is enough to get started without a paid plan.

Usage

/lorebook <universe name> [count | scope: <sub-domain>] [spoiler-light]

Examples:

/lorebook Berserk
/lorebook The Stormlight Archive 100
/lorebook Elden Ring, scope: base game
/lorebook Mistborn, spoiler-light

The entry count is derived automatically from the size of the universe unless you give an explicit count or scope. Defaults are full spoilers and all canon.

Output

The final lorebook is written to output/<slug>.lorebook.json:

The lorebook is in a standard json format any you should be able to import it into SillyTavern, Marinara Engine, or whatever else you're using.

Intermediate working files (entity manifest, raw scraped pages, hand-authored intermediate entries) are written to research/<slug>/ for traceability and are not needed after the build.

How it works

See SKILL.md for the full phased pipeline (source discovery, entity manifest, research, writing, build, QA) and the design decisions behind it.

Repo layout

.claude/skills/lorebook/
  SKILL.md                       the /lorebook skill definition
  scripts/build_lorebook.py      intermediate -> final JSON builder + validator
  references/format.md           JSON format spec (intermediate + final)
  references/writing-style.md    entry content + keyword rules
  examples/                       sample intermediate + final lorebook
research/<slug>/                  per-universe working files (created at runtime, gitignored)
output/<slug>.lorebook.json       final deliverables (created at runtime)
WORKFLOW.md                       pipeline documentation
Ver o original no GitHub

Bom para

Relacionados