Comix Source Plugin for Mugiwara

A built-in source plugin for Mugiwara manga reader that enables importing manga from Comix.

Status

This is an internal built-in plugin that comes pre-installed with Mugiwara. It is not published to NPM and is automatically registered when the server starts.

Features

• Search: Search for manga by title with pagination
• Metadata: Automatic fetching of manga metadata including:
  • Title and alternative titles
  • Description
  • Author and artist information
  • Cover image
  • Genres and tags
  • Publication status and type (Manga/Manhwa/Manhua/Webtoon)
• Chapters: Browse and download all chapters with:
  • Full chapter listing (all pages fetched, pagination handled internally)
  • Deduplication of multi-group duplicate chapter numbers (prefers official scans, then highest votes)
  • Page-by-page streaming download
  • Automatic retry on failures

Configuration

The plugin can be configured via the Mugiwara server configuration:

interface ComixConfig {
  // Base URL for Comix (default: "https://comix.to")
  baseUrl?: string;

  // Request timeout in milliseconds (default: 30000)
  requestTimeout?: number;

  // User agent for requests (default: "Mugiwara/1.0.0")
  userAgent?: string;

  // Number of search results per page (default: 20)
  searchLimit?: number;

  // Number of chapters fetched per chapter-list page (default: 20)
  chapterPageSize?: number;
}

Usage

Since this is a built-in plugin, you don't need to install anything. The source is automatically available with the ID "comix".

API Endpoints

Search for Manga

curl "http://localhost:3000/api/library/sources/comix/search?q=One%20Piece"

Get Manga Details

curl "http://localhost:3000/api/library/sources/comix/manga/{hid}"

Where {hid} is the Comix manga hid (e.g. n8we, returned by search).

Import Manga

curl -X POST "http://localhost:3000/api/library/import" \
  -H "Content-Type: application/json" \
  -d '{
    "sourceId": "comix",
    "sourceMangaId": "n8we",
    "storageBackendId": "default",
    "downloadAll": true
  }'

Programmatic Usage

import createComixSource from "@mugiwara/source-comix";

const source = createComixSource({ config: {} });

await source.initialize({});

const results = await source.search("Solo Leveling");
const manga = await source.getMangaDetails(results[0].sourceMangaId);

for await (const page of source.downloadChapter(
  manga.chapters[0].sourceChapterId,
)) {
  // page.data contains the image buffer
  // page.pageNumber is the page number
  // page.contentType is the MIME type
}

Cloudflare & Crypto

Comix is fronted by Cloudflare but does not present an interactive challenge to plain HTTP fetches. However, the /api/v1/* endpoints enforce an application-layer token: every GET on a signed path must carry a _ token minted from the request's canonical form, and encrypted responses are returned as {"e":"..."} envelopes.

This plugin ships a pure-TS port of the reversed frontend protocol (src/crypto.ts, tables resolved via src/tables.ts):

• makeCrypto(tables) builds signToken(canonical), which reproduces browser-minted _ tokens byte-identically (three chained S-box/XOR stages with keys k367/k432/k55, seeds 189/133/32), plus decryptEnvelope(e) for encrypted responses.

No headless browser is needed at runtime. The S-boxes/keys are baked into a specific frontend build (tl03m9, asset hash 35595e3de3c99889c1aa70); the client detects the live build hash from already-fetched page HTML and uses the pinned src/tables.json while it matches. On-demand derivation from a fresh secure-*.js bundle is designed but not yet implemented (handoff/02-crypto-spec.md §7, tryDeriveTables contract in src/tables.ts), so if Comix redeploys with rotated tables, signed requests will fail with a ComixTablesStaleError naming both the observed and expected builds — then re-snapshot (runbook: handoff/05-re-tooling.md) and refresh src/tables.json.

Endpoints

• Search: GET /api/v1/manga?keyword=...&limit=...&page=...&content_rating[]=... (plaintext, signed)
• Manga details: GET /title/{hid} HTML → embedded initial-data JSON (no token)
• Chapter list: GET /api/v1/manga/{hid}/chapters?page=...&limit=...&order[number]=desc (encrypted, signed)
• Chapter pages: GET /api/v1/chapters/{id} (encrypted, signed) → pages.items[].url
• Page images: direct GET on the per-chapter image host, requires Referer: https://comix.to/ (else 403). Hosts vary per chapter — URLs are always taken verbatim from the API payload, never hardcoded.

Retry Logic

The plugin implements the following retry behavior for page downloads:

• 100ms delay between page downloads
• Exponential backoff on retry (1s, 2s, 3s)
• 30-second timeout on API requests
• Maximum 3 retries per page

License

MIT

Disclaimer

This plugin is not affiliated with Comix. Please respect their terms of service when using this plugin.