feat: R2 поддержки Cloud — запись страниц и вложений

CHANGELOG.en.md

@@ -6,6 +6,29 @@ All notable changes to the Confluence Page Exporter tool are documented in this
 
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/).
 
+## [Unreleased]
+
+### Added
+
+- Write support for Confluence Cloud: `upload update` / `upload create` /
+  `upload merge` and all attachment operations now work on Cloud — Cloud
+  support is complete. Pages are created and updated via REST API v2 (the
+  numeric space id is resolved from the key automatically; a version conflict
+  arrives as 409 and is handled as before), attachments are uploaded through
+  the v1 endpoints Atlassian kept on Cloud. Verified end-to-end against a
+  live Cloud site, including both-sides conflict detection.
+
+### Fixed
+
+- Removed the phantom "changed on server" in `compare` for pages whose macros
+  were authored without service attributes: Confluence adds
+  `ac:schema-version` to macros on save, and the Cloud editor stamps
+  `local-id`/`ac:local-id` onto elements — the normalizer now ignores these
+  attributes during comparison (uploaded content is not affected). The
+  canonicalization epoch is bumped to 3: hashes stored in markers are
+  recomputed on each page's next sync; until then local-edit detection falls
+  back to file modification times, once.
+
 ## [2.17.0] — 2026-07-02
 
 ### Added

CHANGELOG.md

@@ -6,6 +6,30 @@
 
 Формат основан на [Keep a Changelog](https://keepachangelog.com/ru/1.0.0/).
 
+## [Unreleased]
+
+### Добавлено
+
+- Поддержка записи на Confluence Cloud: `upload update` / `upload create` /
+  `upload merge` и все операции с вложениями теперь работают на Cloud —
+  поддержка Cloud стала полной. Страницы создаются и обновляются через REST
+  API v2 (числовой идентификатор пространства резолвится из ключа
+  автоматически, конфликт версий приходит как 409 и обрабатывается как
+  раньше), вложения загружаются через сохранившиеся на Cloud v1-эндпоинты.
+  Проверено сквозным тестом на живом Cloud-сайте, включая детект конфликтов
+  «изменено с обеих сторон».
+
+### Исправлено
+
+- Убран фантомный «изменено на сервере» в `compare` для страниц, чьи макросы
+  были созданы без служебных атрибутов: Confluence при сохранении добавляет
+  макросам `ac:schema-version`, а редактор Cloud штампует на элементы
+  `local-id`/`ac:local-id` — нормализатор теперь игнорирует эти атрибуты при
+  сравнении (на отправляемый контент это не влияет). Эпоха канонизации
+  поднята до 3: сохранённые в маркерах хэши пересчитаются при следующей
+  синхронизации каждой страницы, до этого детект локальных правок
+  одноразово опирается на время изменения файлов.
+
 ## [2.17.0] — 2026-07-02
 
 ### Добавлено

README.en.md

@@ -25,7 +25,7 @@ The tool supports:
   - attachments as separate files
   - a marker file `.id<pageId>_<version>` (its body is JSON with the original title and space key) for stable page identification and version tracking
 - Authentication modes: `--auth-type onprem` and `--auth-type cloud`; by default the type is auto-detected from `--base-url` (`*.atlassian.net` hosts → `cloud`)
-- Confluence Cloud support (REST API v2) — read-only for now: `download update`/`download merge`, `compare`, and the `confluence_ping` / `confluence_get_page_content` MCP tools. Write operations (upload, attachments) on Cloud fail with a clear error and arrive in upcoming releases
+- Full Confluence Cloud support (REST API v2 + the v1 endpoints Atlassian kept): every command — download, upload, compare, attachments — works the same against Server/DC and Cloud; Cloud's numeric space ids are transparently resolved from keys
 - Multi-layered configuration with priority: CLI > environment variables > file > default value
 - Global `--verbose` flag for detailed (debug-level) output
 - Dry-run support where applicable

README.md

@@ -25,7 +25,7 @@
   - вложения как отдельные файлы
   - файл-маркер `.id<pageId>_<version>` (в теле — JSON с оригинальным заголовком и ключом пространства) для стабильной идентификации страницы и отслеживания версии
 - Режимы аутентификации: `--auth-type onprem` и `--auth-type cloud`; по умолчанию тип определяется автоматически по `--base-url` (хосты `*.atlassian.net` → `cloud`)
-- Поддержка Confluence Cloud (REST API v2) — пока только чтение: `download update`/`download merge`, `compare`, MCP-инструменты `confluence_ping` и `confluence_get_page_content`. Операции записи (upload, вложения) на Cloud вернут понятную ошибку и появятся в следующих релизах
+- Полная поддержка Confluence Cloud (REST API v2 + сохранившиеся v1-эндпоинты): все команды — download, upload, compare, вложения — работают одинаково на Server/DC и Cloud; числовые идентификаторы пространств Cloud прозрачно резолвятся из ключей
 - Многоуровневая конфигурация с приоритетом: CLI > переменные окружения > файл > значение по умолчанию
 - Глобальный параметр `--verbose` для подробного (debug-level) вывода
 - Поддержка dry-run там, где применимо

docs/mcp/agent-instructions.md

@@ -10,7 +10,7 @@ These same instructions are also published as `docs/mcp/agent-instructions.md` i
 
 - **Sandbox root.** The server was started with `--root-dir <path>`. Every path you pass to a tool is resolved relative to that root; absolute paths are accepted only when they lie inside it. Outside paths fail with `OUT_OF_SANDBOX`. **You cannot widen this sandbox** — it is set at server startup and ignores env vars and config files.
 - **Read-only mode.** If the server was started with `--read-only`, all `confluence_upload_*` tools refuse with `READ_ONLY_VIOLATION`. `confluence_download_*`, `confluence_compare`, `confluence_ping`, and `confluence_get_page_content` remain available.
-- **Confluence Cloud (read-only phase).** Against a Cloud site (auto-detected from a `*.atlassian.net` base URL, or `AuthType=cloud`) the download/compare/ping/get-page-content tools work via the Cloud v2 API; all `confluence_upload_*` tools currently fail with `NOT_SUPPORTED` — Cloud write support has not shipped yet. Do not retry; report it to the user.
+- **Confluence Cloud.** Cloud sites (auto-detected from a `*.atlassian.net` base URL, or `AuthType=cloud`) are fully supported — every tool works the same as against Server/Data Center. No Cloud-specific behaviour to account for.
 - **Space keys (multi-space).** A default `spaceKey` may be configured server-side, and most tools accept an optional `spaceKey` that overrides it. For an **already-synced tree you normally don't need to pass `spaceKey` at all** — each page records its space locally (in the `.id` marker) and the server is the authority. Passing a `spaceKey` that contradicts a page's real space makes the upload tools **error** rather than silently mis-place pages. New child pages are always created in their parent's space (a Confluence tree lives in exactly one space). See "Working with multiple spaces" below.
 
 ## Local layout
@@ -64,7 +64,7 @@ Every tool returns one of two shapes:
 | `INVALID_ARGS` | You passed contradictory or missing arguments. Read `error` and try again. |
 | `OUT_OF_SANDBOX` | A path argument resolved outside `--root-dir`. Use a path inside the sandbox. |
 | `READ_ONLY_VIOLATION` | You called an upload tool on a read-only server. Stop trying. |
-| `NOT_SUPPORTED` | The operation is not available for this deployment — currently any write against Confluence Cloud (read-only Cloud phase). Stop trying and tell the user. |
+| `NOT_SUPPORTED` | The operation is not available for this deployment. Stop trying and tell the user. |
 | `AUTH_FAILED` | 401/403 from Confluence or a local FS auth error. Report to the user. |
 | `PAGE_NOT_FOUND` | Page ID/title doesn't exist on the server. Re-check inputs. |
 | `NETWORK_ERROR` | `HttpRequestException` with no status — DNS/TCP/SSL. The server already retried up to 3× with backoff, so this means all retries failed. Call `confluence_ping` to confirm the server can reach Confluence. |

src/ConfluencePageExporter/Services/ConfluenceApiHelpers.cs

@@ -1,4 +1,5 @@
 using System.Net;
+using System.Net.Http.Headers;
 
 namespace ConfluencePageExporter.Services;
 
@@ -42,4 +43,68 @@ internal static async Task EnsureSuccessAsync(HttpResponseMessage response, stri
     /// </summary>
     internal static string EscapeCql(string value) =>
         value.Replace("\\", "\\\\").Replace("\"", "\\\"");
+
+    private const string DefaultAttachmentMediaType = "application/octet-stream";
+
+    /// <summary>
+    /// Minimal extension → media-type map for the attachment upload path.
+    /// Confluence infers an attachment's media type from the multipart part's
+    /// Content-Type; when we send none it falls back to the filename
+    /// extension, which has no answer for extensionless files. This covers
+    /// the common attachment kinds so a new attachment keeps its proper type,
+    /// with <see cref="DefaultAttachmentMediaType"/> as the catch-all.
+    /// </summary>
+    private static readonly IReadOnlyDictionary<string, string> ExtensionMediaTypes =
+        new Dictionary<string, string>(StringComparer.OrdinalIgnoreCase)
+        {
+            [".png"] = "image/png",
+            [".jpg"] = "image/jpeg",
+            [".jpeg"] = "image/jpeg",
+            [".gif"] = "image/gif",
+            [".bmp"] = "image/bmp",
+            [".webp"] = "image/webp",
+            [".svg"] = "image/svg+xml",
+            [".pdf"] = "application/pdf",
+            [".txt"] = "text/plain",
+            [".csv"] = "text/csv",
+            [".xml"] = "application/xml",
+            [".json"] = "application/json",
+            [".html"] = "text/html",
+            [".htm"] = "text/html",
+            [".zip"] = "application/zip",
+            [".drawio"] = "application/vnd.jgraph.mxfile",
+        };
+
+    /// <summary>
+    /// Resolves the media type to stamp on an uploaded attachment part.
+    /// Prefers the caller-supplied type (on update: the existing attachment's
+    /// stored server type, so the update never changes it), then the extension
+    /// map, and finally <see cref="DefaultAttachmentMediaType"/> for
+    /// extensionless/unknown files.
+    /// </summary>
+    private static string ResolveAttachmentContentType(string fileName, string? preferredMediaType)
+    {
+        if (!string.IsNullOrWhiteSpace(preferredMediaType))
+            return preferredMediaType;
+
+        var ext = Path.GetExtension(fileName);
+        if (!string.IsNullOrEmpty(ext) && ExtensionMediaTypes.TryGetValue(ext, out var mapped))
+            return mapped;
+
+        return DefaultAttachmentMediaType;
+    }
+
+    /// <summary>
+    /// Sets the multipart file part's Content-Type from
+    /// <see cref="ResolveAttachmentContentType"/>, falling back to
+    /// <see cref="DefaultAttachmentMediaType"/> if the resolved value is not a
+    /// valid media-type header (e.g. an odd server-reported value).
+    /// </summary>
+    internal static void ApplyAttachmentContentType(ByteArrayContent part, string fileName, string? preferredMediaType)
+    {
+        var resolved = ResolveAttachmentContentType(fileName, preferredMediaType);
+        if (!MediaTypeHeaderValue.TryParse(resolved, out var parsed))
+            MediaTypeHeaderValue.TryParse(DefaultAttachmentMediaType, out parsed);
+        part.Headers.ContentType = parsed;
+    }
 }