[Phase B2] Generalize SourceFetcher beyond note.com (Zenn, Qiita, RSS, HTML)

[terisuke]

Tracked under ADR 0002 — Phase B. Detail: implementation plan §B2.

Problem

Style derivation is locked to note.com. The owner publishes under multiple identities:

• Terisuke (本人): note.com/cor_instrument, cor-jp.com/blog
• Cloudia (キャラクター): zenn.dev/cloudia, qiita.com/Cloudia_Cor_Inc

Every persona must be able to derive its style guide from its own platforms.

Scope

New interface in internal/domain/source:

type Fetcher interface {
    FetchProfile(ctx context.Context, ref Ref) (*ProfileSnapshot, error)
    FetchArticle(ctx context.Context, ref Ref) (*ArticleSnapshot, error)
    FetchList(ctx context.Context, ref Ref, limit int) ([]ArticleSnapshot, error)
}

Concrete implementations under internal/infrastructure/source/:

• note/ — extracted from existing internal/infrastructure/note/fetcher.go. The note.com host check moves into this fetcher only.
• zenn/ — public articles + /{user}/feed. Robots.txt + 1 req/sec rate limit.
• qiita/ — public REST API (no auth needed for read-only public posts) + HTML fallback.
• rss/ — generic RSS reader for Astro/Jekyll/Hugo blogs (cor-jp.com once its RSS is verified).
• html/ — generic semantic-content extractor (last resort).

Each fetcher has its own User-Agent string and rate-limit policy.

Acceptance criteria

• Scenario test fetches one article from each of {note, zenn, qiita, rss} and writes tmp/source_fetch/{name}.json.
• application and handlers packages no longer reference note.com directly; routing decisions live in the persona registry or fetcher dispatcher.
• Cloudia's Zenn and Qiita articles are fetched in a scenario test; both produce non-empty ArticleSnapshot with code blocks preserved.
• Each fetcher has unit tests using httptest.Server (no real network).

Out of scope

• Authenticated fetching (private articles, drafts).
• Pagination beyond the first page.

Dependencies

• B1 should land first so that the persona registry can declare which fetcher to use per source.

[terisuke]

Implemented and merged via PR #52. Validation is recorded in docs/validation/issue-22-source-fetchers-2026-05-02.md. The source router now supports note, Zenn RSS with all=1, Qiita API v2 user items, RSS/Atom, and generic HTML.

[terisuke]

検証不足がありました。前回の #22 は各ソース1件の疎通確認に留まっており、Cloudia/Zenn/Qiita/Corブログの過去記事を本文つきで取得できることを証明できていません。ユーザー別の履歴取得・Zenn本文取得・CorブログMarkdown/RSS取得を再検証するため、このIssueを再オープンします。

[terisuke]

PR #55 を develop にマージし、#22 の検証不足を補正しました。2026-05-02 時点の実サイトシナリオで zenn:cloudia は7件・Qiita Cloudia_Cor_Inc は12件・Cor RSS は10件・Cor GitHub Markdown は10件を取得確認済みです。RSSは概要のみのため、Corブログの本文スタイル分析は github:Cor-Incorporated/corsweb2024/src/content/blog/ja を正とします。go test ./... も通過済みです。