endfield-mcp/dist/data/characterProfiles.js

MCP Server for Arknights: Endfield (TypeScript / stdio + Streamable HTTP)

/**
 * Character profile (archives) and voice-line reader.
 *
 * This is the **fan-creation core** of the character domain — archives
 * give the background-story text a writer needs, voice lines give the
 * speech patterns. The numeric fields (profession, rarity, weaponType)
 * live in characters.ts:getCharacterInfo for the gameplay-curious.
 *
 * Both profileRecord and profileVoice live inside CharacterTable.json
 * (unlike PRTS-MCP where they're in separate handbook/charword tables),
 * so this module reuses characters.ts's table loader and resolver.
 *
 * ## Text cleaning
 *
 * Endfield's profile text contains rich-text tags like `<@profile.key>`
 * (key markers) and `</>` (close markers). These are display hints, not
 * content — we strip them so the LLM gets clean prose. Voice lines are
 * typically plain text without tags.
 */
import { resolveCharacterEntry } from "./characters.js";
import { resolveText, } from "./texts.js";
// ---------------------------------------------------------------------------
// Text cleaning
// ---------------------------------------------------------------------------
/**
 * Strip Endfield rich-text markup from profile/voice text.
 *
 * Two opening-tag families exist in the data (both closed by `</>`):
 *   <@profile.key>...</>   — key/highlight span (98+ distinct tags)
 *   <#ba.consume>...</>    — status/effect span (98 distinct tags, 870+ values)
 *
 * Both are display hints, not content. We strip them so the LLM gets
 * clean prose. The close tag `</>` is shared across both families.
 */
export function cleanProfileText(text) {
    // Drop opening tags: <@...> and <#...> (the two Endfield tag families).
    let out = text.replace(/<[#@][a-zA-Z0-9_.]+>/g, "");
    // Drop close tags </>
    out = out.replace(/<\/>/g, "");
    return out.trim();
}
// ---------------------------------------------------------------------------
// Public API
// ---------------------------------------------------------------------------
/**
 * Get a character's archive sections (background story text).
 *
 * Returns 3 sections per character (basic profile / personnel summary /
 * archive material), each with a title and full prose body. Text is
 * cleaned of rich-text tags and localized to the requested language.
 *
 * Returns null if the character is not found or has no archive data.
 */
export function getCharacterArchives(idOrName, lang = "CN") {
    const resolved = resolveCharacterEntry(idOrName, lang);
    if (resolved === null)
        return null;
    const records = resolved.entry
        .profileRecord;
    if (!records || records.length === 0)
        return null;
    return records.map((r) => ({
        recordId: r.recordID ?? "",
        title: resolveText(r.recordTitle, lang, r.recordID ?? ""),
        text: cleanProfileText(resolveText(r.recordDesc, lang, "")),
    }));
}
/**
 * Get a character's voice lines (speech text with trigger conditions).
 *
 * Returns ~55 lines per character, each with an index, a trigger title
 * (e.g. "行动准备1", "编入队伍1"), and the spoken text. Localized to
 * the requested language.
 *
 * Returns null if the character is not found or has no voice data.
 */
export function getCharacterVoices(idOrName, lang = "CN") {
    const resolved = resolveCharacterEntry(idOrName, lang);
    if (resolved === null)
        return null;
    const voices = resolved.entry
        .profileVoice;
    if (!voices || voices.length === 0)
        return null;
    return voices.map((v) => ({
        index: v.voiceIndex ?? 0,
        title: resolveText(v.voiceTitle, lang, v.voId ?? ""),
        text: cleanProfileText(resolveText(v.voiceDesc, lang, "")),
    }));
}
//# sourceMappingURL=characterProfiles.js.map