chore: promote develop docs roadmap to main

README.md

@@ -4,6 +4,8 @@ Note記事のURLまたはユーザー名を入力し、ローカルLLMが新し
 
 今後の大規模改修方針は [ADR 0001](docs/adrs/0001-three-phase-local-article-generation.md) と [3フェーズ実装計画](docs/implementation-plans/three-phase-local-article-generation.md) に整理しています。単一フォームで記事を生成する方式から、文体分析、記事条件の一問一答、記事生成と評価の3フェーズへ移行します。
 
+ADR 0001 を踏まえた次の進化方針は [ADR 0002 — Multi-Persona, Multi-Format Article Generation](docs/adrs/0002-multi-persona-multi-format-extension.md) と [Multi-persona / multi-format 実装計画](docs/implementation-plans/multi-persona-multi-format.md) にまとめています。てりすけ本人と架空キャラ「宇宙野クラウディア」を別人格として扱い、note / cor-jp.com ブログ / Zenn / Qiita / ホームページHTML を切り替え可能にします。
+
 実装 issue と ADR の対応、各層の責務、テスト条件は [Issue and ADR guardrails](docs/implementation-plans/issue-adr-guardrails.md) にまとめています。
 
 ## 主な機能

docs/adrs/0001-three-phase-local-article-generation.md

@@ -4,7 +4,7 @@ Date: 2026-05-01
 
 ## Status
 
-Accepted
+Accepted. Extended by [ADR 0002](0002-multi-persona-multi-format-extension.md), which adds the orthogonal axes of author persona and output format on top of the three-phase pipeline. The phases described below remain authoritative; the registries and strategies introduced by ADR 0002 plug into them rather than replacing them.
 
 ## Context
 

docs/adrs/0002-multi-persona-multi-format-extension.md

@@ -0,0 +1,233 @@
+# ADR 0002: Multi-Persona, Multi-Format Article Generation
+
+Date: 2026-05-02
+
+## Status
+
+Accepted. Extends and partially supersedes [ADR 0001](0001-three-phase-local-article-generation.md). The three-phase workflow (style analysis → interview → draft) is preserved. This ADR adds two orthogonal axes — **author persona** and **output format** — and reshapes UX, persistence, and prompt construction accordingly.
+
+## Context
+
+ADR 0001 produced a working three-phase pipeline for `cor_instrument` (Terisuke) on note.com. Real usage by the project owner exposed three structural gaps:
+
+1. **The owner publishes under multiple identities and to multiple platforms.**
+   - Terisuke (本人): [note.com/cor_instrument](https://note.com/cor_instrument), [cor-jp.com/blog](https://cor-jp.com/blog/) — reflective entrepreneur essays mixed with technical experience reports. First person 「僕」「私」, narrative arcs ("～した話", "～てしまった件"), philosophical framing.
+   - Cloudia / 宇宙野クラウディア (架空キャラクター, also operated by the owner): [zenn.dev/cloudia](https://zenn.dev/cloudia), [qiita.com/Cloudia_Cor_Inc](https://qiita.com/Cloudia_Cor_Inc) — character-branded technical tutorials. Hakata-ben (博多弁) flavour, exclamation-driven titles ("クラウディア流！…", "AI探検記【前編】〜最強のお助けAIを探せ！〜"), strong code-block density.
+   - Treating these as one author profile contaminates the style guide and produces drafts that read as neither voice.
+
+2. **The output format is hard-coded to "note.com paste-ready Markdown article".**
+   - `internal/domain/article/draft.go:25-35` requires a `# title` first line and rejects code fences.
+   - `internal/application/draft/prompt.go:12-51` injects "Noteにそのまま貼り付けられる日本語Markdown記事" into every system message.
+   - `internal/infrastructure/note/fetcher.go:234-246` rejects any host other than `note.com` for source ingestion.
+   - HTML homepage sections, Zenn/Qiita Markdown articles (which carry frontmatter), and Astro-blog posts cannot be produced today.
+
+3. **The current UX does not feel like working with a writing partner.**
+   - The interview is a forward-only, edit-locked Q&A; users cannot revise an earlier answer.
+   - The completed brief is rendered as `JSON.stringify` (`static/js/script.js:178`); accumulated knowledge is not legible.
+   - The draft textarea is `readonly` (`static/index.html:153`); there is no per-section regenerate.
+   - The full-screen spinner blocks the page during every LLM call; nothing streams.
+   - There is no visible history of past style profiles or past sessions; reuse is invisible.
+
+The user's stated comparators are Claude, ChatGPT, and Gemini chat UIs — continuous transcripts with editable history, streaming output, and a side-pinned working artifact.
+
+## Decision
+
+Add two orthogonal first-class concepts to the domain and let the rest of the system depend on them via strategy boundaries.
+
+### Persona
+
+A `Persona` is a named, reusable bundle of:
+
+- one or more `AuthorSource` adapters that supply training material,
+- one canonical `WritingStyleGuide` (regenerable when sources change),
+- a default `OutputFormat` (overridable per article),
+- a default question-set variant for the interview phase,
+- optional preset values for the brief (preferred first person, signature opening patterns, recurring themes, anti-patterns).
+
+Two personas ship pre-loaded:
+
+| Persona | Display name | Sources | Default format | Voice notes |
+|---|---|---|---|---|
+| `terisuke` | てりすけ | `note.com/cor_instrument`, `cor-jp.com/blog/*` | `note_article` | 一人称「僕」/「私」、内省＋実体験ナラティブ、起業・キャリア・AI駆動開発、「～した話」「～てしまった件」 |
+| `cloudia` | 宇宙野クラウディア | `zenn.dev/cloudia`, `qiita.com/Cloudia_Cor_Inc` | `zenn_article` | 一人称「クラウディア」/「うち」、博多弁混じり、感嘆符・【前編】等の装飾、AI/JS/Pythonチュートリアル、感情的訴求 (「劇的に」「最強の」) |
+
+Personas are user-extensible. Adding a third persona requires only registering it (no code changes inside the prompt builder).
+
+### OutputFormat
+
+An `OutputFormat` is a typed strategy with:
+
+- a Markdown/HTML template surface (frontmatter, heading rules, code-fence policy, length envelope),
+- a validator (e.g., Zenn requires frontmatter; note rejects code fences; homepage sections forbid `# title`),
+- a default question-set extension (technical formats add "対象スタック", "実行環境", "想定読者の前提知識"; narrative formats add "導入エピソード", "結末で読者に届けたい感情"),
+- a system-prompt fragment merged into the draft prompt (instead of a single hard-coded fragment).
+
+Formats shipped in v2:
+
+| Format | Validator highlights | Source/target |
+|---|---|---|
+| `note_article` | Existing rules: starts with `# `, no code fences, `ですます調` default | note.com paste |
+| `markdown_blog` | Frontmatter optional, allows code fences, `## ` first heading allowed (Astro blogs at `cor-jp.com/blog`) | cor-jp.com |
+| `zenn_article` | Frontmatter required (`title`, `emoji`, `type`, `topics`, `published`), code fences allowed, technical tone bias | zenn.dev/cloudia |
+| `qiita_article` | Frontmatter required (`title`, `tags`), code blocks expected, technical tone bias | qiita.com/Cloudia_Cor_Inc |
+| `homepage_section` | HTML output, no `# title`, semantic sectioning, CTA placement guidance | static HTML embed |
+
+The two axes compose: any persona can produce any format. The application layer rejects nonsensical combinations only when the persona explicitly excludes a format (configurable, not hard-coded).
+
+### UX direction
+
+The single-page form becomes a **conversation-first workspace**:
+
+- **Left rail**: persona switcher, project history, "previous style guides" picker.
+- **Centre**: chat-style transcript. Past answers are clickable and editable, which forks a new session draft from that point. Deep-dive questions render with the parent question quoted as context.
+- **Right rail (artifact panel)**: live brief card and live draft preview, replacing the current JSON dump. Updates as the conversation progresses.
+- **Streaming**: SSE for follow-up generation and draft generation. The full-screen spinner is removed.
+- **Draft editor**: editable in-place, with "regenerate this section" (selection-based) and "copy to clipboard" buttons for both raw Markdown and rendered HTML output.
+
+The static-JS prototype is preserved as a fallback. The new UX can ship as a progressive overlay (no SPA-framework rewrite required for v2; if desired, a later ADR can introduce one).
+
+### Persistence direction
+
+The flat `data/workflow_store.json` snapshot is replaced by a SQLite-backed store with explicit aggregates:
+
+- `personas` — id, display name, default format, source bundle, current canonical guide id.
+- `author_sources` — persona-scoped fetched articles with raw text snapshot for re-derivation.
+- `writing_style_guides` — versioned per persona; previous versions retained.
+- `projects` — a writing project (e.g., "Q3 cor-jp blog series") that owns multiple articles.
+- `articles` — one target deliverable: persona id, output format, brief id, draft history.
+- `brief_sessions`, `brief_answers` — unchanged in shape, gain a `parent_answer_id` for fork-on-edit.
+- `drafts` — versioned per article with score history.
+
+Acceptance criterion: any prior session can be reopened, its accumulated context shown as a transcript, and a new draft regenerated from any point in history.
+
+This subsumes Issue [#14](https://github.com/terisuke/note_maker/issues/14) (queryable database). Issue [#14](https://github.com/terisuke/note_maker/issues/14) is kept open as the umbrella tracker; the SQLite migration becomes its acceptance.
+
+### Source acquisition direction
+
+`SourceFetcher` becomes a strategy interface with concrete implementations:
+
+- `NoteFetcher` — existing `note.com` page + RSS (kept).
+- `ZennFetcher` — public articles (`https://zenn.dev/{user}/articles/{slug}`) and the user's article list.
+- `QiitaFetcher` — public REST (no auth needed for public posts; respect rate limits).
+- `GenericRSSFetcher` — for Astro/Jekyll/Hugo blogs that expose RSS (cor-jp.com is included here once its RSS is confirmed).
+- `HTMLFetcher` — fallback for arbitrary HTML pages with semantic-content extraction.
+
+Per-fetcher rate-limit and User-Agent policy live alongside each adapter.
+
+## Domain Model Changes
+
+New domain types under `internal/domain`:
+
+- `persona` package
+  - `Persona` (id, display_name, sources, default_format, default_questions_extension)
+  - `PersonaRegistry` (in-memory + persisted, seeded with `terisuke` and `cloudia`)
+- `format` package
+  - `OutputFormat` (id, validator, template fragment, default_question_extension)
+  - `FormatRegistry`
+  - `Validator` interface; per-format implementations (e.g., `NoteValidator`, `ZennValidator`)
+- `domain/article`
+  - `Draft` gains a `format_id`; `NewDraft` dispatches to the format's validator instead of hard-coded rules.
+- `domain/brief`
+  - `ArticleBriefSession` gains `persona_id`, `output_format_id`, `parent_answer_id`.
+  - `FixedQuestions` becomes a base list extended by the persona and the format.
+- `domain/project` (new)
+  - `Project`, `Article` aggregates as defined in the persistence section.
+
+## Application Service Changes
+
+- `AnalyzeAuthorStyleService` accepts a `persona_id` and persists the resulting guide as a new version under that persona; previous versions are preserved.
+- `InterviewService` consults the active persona and format to assemble the question list before the first question.
+- `GenerateDraftService` resolves the prompt template fragment from the format's strategy and merges persona-specific tone hints.
+- New `RegenerateSectionService` accepts a draft id, a section selector (heading anchor or character range), the brief, and the persona+format; returns a candidate replacement for human review.
+- New `StreamingDraftService` produces SSE chunks for the draft phase.
+
+## Infrastructure Changes
+
+- `internal/infrastructure/repository/sqlite` — new package implementing every repository interface; the JSON file repository becomes an export/import utility for portability.
+- `internal/infrastructure/source/{note,zenn,qiita,rss,html}` — per-source fetchers behind a common interface in `internal/domain/source` (or kept under `infrastructure` and bound by interface in `domain/persona`).
+- `internal/infrastructure/llamacpp` — gains streaming (SSE) support; existing non-streaming path retained for tests.
+
+## API Changes
+
+Additions:
+
+- `GET /api/personas` / `POST /api/personas` / `PATCH /api/personas/{id}` — persona CRUD.
+- `GET /api/formats` — read-only registry of available formats.
+- `POST /api/projects` / `GET /api/projects` / `GET /api/projects/{id}` — project management.
+- `GET /api/sessions/{id}/transcript` — chat-style transcript including parent links.
+- `POST /api/sessions/{id}/answers/{answer_id}/edit` — fork-on-edit for past answers.
+- `POST /api/drafts/{id}/regenerate-section` — section-level regenerate.
+- `POST /api/author-style/analyze`, `POST /api/drafts` — gain `Accept: text/event-stream` for streaming.
+
+Existing `/api/generate` remains a compatibility facade.
+
+## Testing Strategy
+
+- Unit tests added per format validator, per persona seed, per fetcher.
+- Integration tests for full transcript edit-and-fork flow.
+- Scenario tests:
+  - `cmd/scenario/persona_terisuke_note` — current behaviour.
+  - `cmd/scenario/persona_terisuke_blog` — new, targets cor-jp blog format.
+  - `cmd/scenario/persona_cloudia_zenn` — new, validates Zenn frontmatter and code-fence presence.
+  - `cmd/scenario/persona_cloudia_qiita` — new.
+- HTTP handler tests added for every endpoint in `internal/handlers/workflow.go` (currently uncovered).
+- Playwright E2E (extends Issue [#13](https://github.com/terisuke/note_maker/issues/13)) covers persona switch, format switch, edit-and-fork, streaming completion, copy-clipboard, regenerate-section.
+
+## Phased Rollout
+
+The full work is broken into four phases tracked by issues. Each phase is independently shippable and behind a UI toggle until ready.
+
+- **Phase A — Conversation UX (1–2 weeks)**
+  - Chat transcript with editable answers, streaming, deep-dive context display, draft editor + per-section regenerate.
+  - Issues: [#17](https://github.com/terisuke/note_maker/issues/17), [#18](https://github.com/terisuke/note_maker/issues/18), [#19](https://github.com/terisuke/note_maker/issues/19), [#20](https://github.com/terisuke/note_maker/issues/20).
+- **Phase B — Multi-persona, multi-format (2–3 weeks)**
+  - Persona registry seeded with `terisuke` and `cloudia`, OutputFormat strategy, source fetcher generalisation, format-aware question sets.
+  - Issues: [#21](https://github.com/terisuke/note_maker/issues/21), [#22](https://github.com/terisuke/note_maker/issues/22), [#23](https://github.com/terisuke/note_maker/issues/23), [#24](https://github.com/terisuke/note_maker/issues/24), [#25](https://github.com/terisuke/note_maker/issues/25).
+- **Phase C — Memory & history (2 weeks, integrates Issue [#14](https://github.com/terisuke/note_maker/issues/14))**
+  - SQLite store with project/article schema, profile/session reuse UI, brief and guide rendered as cards.
+  - Issues: [#26](https://github.com/terisuke/note_maker/issues/26), [#27](https://github.com/terisuke/note_maker/issues/27), [#28](https://github.com/terisuke/note_maker/issues/28).
+- **Phase D — Quality & ops**
+  - Handler test coverage, Issue [#11](https://github.com/terisuke/note_maker/issues/11) (style threshold), Issue [#13](https://github.com/terisuke/note_maker/issues/13) (Playwright), Issue [#15](https://github.com/terisuke/note_maker/issues/15) (desktop packaging) follow-up.
+  - Issue: [#29](https://github.com/terisuke/note_maker/issues/29).
+
+Recommended order: A → C → B → D. Phase C is sequenced before B because the persona registry needs durable storage to be useful; running B on the JSON store would force a second migration.
+
+## Tracked issues
+
+Filed 2026-05-02 as part of the PR that introduced this ADR.
+
+- A1 — [#17](https://github.com/terisuke/note_maker/issues/17) Refactor interview UI to chat-style transcript with editable past answers
+- A2 — [#18](https://github.com/terisuke/note_maker/issues/18) Stream LLM responses via SSE for follow-up and draft generation
+- A3 — [#19](https://github.com/terisuke/note_maker/issues/19) Editable draft Markdown + per-section regenerate API
+- A4 — [#20](https://github.com/terisuke/note_maker/issues/20) Surface deep-dive question rationale in prompt and UI
+- B1 — [#21](https://github.com/terisuke/note_maker/issues/21) Introduce Persona and OutputFormat domain concepts (registry + strategy)
+- B2 — [#22](https://github.com/terisuke/note_maker/issues/22) Generalize SourceFetcher beyond note.com (Zenn, Qiita, RSS, HTML)
+- B3 — [#23](https://github.com/terisuke/note_maker/issues/23) Format-specific prompt templates and draft validators (note / markdown_blog / zenn / qiita / homepage_section)
+- B4 — [#24](https://github.com/terisuke/note_maker/issues/24) Seed persona library with `terisuke` and `cloudia` profiles
+- B5 — [#25](https://github.com/terisuke/note_maker/issues/25) Format- and persona-aware fixed question sets
+- C1 — [#26](https://github.com/terisuke/note_maker/issues/26) Replace JSON store with SQLite-backed schema (extends [#14](https://github.com/terisuke/note_maker/issues/14))
+- C2 — [#27](https://github.com/terisuke/note_maker/issues/27) Persona / past-session picker UI
+- C3 — [#28](https://github.com/terisuke/note_maker/issues/28) Render brief and style guide as human-readable cards
+- D1 — [#29](https://github.com/terisuke/note_maker/issues/29) HTTP handler tests for `internal/handlers/workflow.go` (currently 0% coverage)
+
+## Consequences
+
+Positive:
+
+- The product can serve both Terisuke (本人) and Cloudia (キャラクター) without contaminating either voice.
+- Adding a new platform (e.g., Substack, dev.to) becomes one fetcher + one format, not a fork.
+- Memory becomes legible: persona library, project history, and brief cards make accumulated context visible.
+- The conversation feel matches the Claude/ChatGPT/Gemini comparator class.
+
+Tradeoffs:
+
+- Domain surface grows. The win is mitigated by registries and strategy boundaries instead of conditionals.
+- SQLite migration is unavoidable; the JSON file becomes export-only.
+- Streaming requires keeping non-streaming paths for tests; carries minor duplication.
+
+## Rejected alternatives
+
+- **One persona with `style_variant` flag.** Rejected because the two voices share *no* tonal substrate; mixing them in one guide demonstrably degrades both. Two distinct guides are operationally cleaner than one guide with branches.
+- **Output format as a free-text instruction in the brief.** Rejected because validators must be code-enforced (Zenn frontmatter, note `# title`, HTML semantics). Free text gives no validator handle.
+- **Single-page rewrite to React/Vue first.** Rejected as premature; the UX wins (chat transcript, streaming, editable draft) are achievable as progressive enhancements. A framework rewrite is worth a later ADR if profiling shows the static prototype slows iteration.
+- **Skip persona model and let the user paste a custom style guide.** Rejected because re-deriving the guide from real articles is the strongest signal we have; manual paste defeats the style-comparator scoring.