feat(observabilidade): tabela worklog_comments + parser HTML + importer MCP

- Schema worklog_comments (id, discussion, parent, datas, staff, campos parseados em JSON)
- Parser HTML tolerante (h2/h3/h4) extrai title, task_ref, duration, work_items,
  files_modified, problems, patterns_text, actions
- Módulo worklog-import com paginação MCP get_discussion_comments
- Helper mcp-client.ts partilhado (gateway MCP JSON-RPC + SSE)
- Dep runtime: node-html-parser

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

api/services/sessions/db.ts

@@ -27,6 +27,31 @@ export interface PatternRecord {
   consecutive_weeks: number
 }
 
+export interface WorklogCommentRecord {
+  id: number
+  discussion_id: number
+  created_at: string
+  staff_id: number | null
+  title: string | null
+  task_ref: string | null
+  duration_sec: number | null
+  work_items: string[]
+  files_modified: string[]
+  problems: { problema: string; solucao: string }[]
+  patterns_text: string[]
+  actions: { tipo: string; descricao: string; prioridade: string | null }[]
+  raw_html: string
+  imported_at: string
+}
+
+export interface WorklogFilters {
+  discussion_id?: number
+  task_ref?: string
+  sinceIso?: string
+  limit?: number
+  offset?: number
+}
+
 export interface SessionsDb {
   upsertSession(meta: SessionMeta): void
   upsertMany(metas: SessionMeta[]): void
@@ -37,6 +62,10 @@ export interface SessionsDb {
   upsertPattern(p: PatternRecord): void
   getPatternsByWeek(week: string): PatternRecord[]
   getConsecutiveWeeks(pattern_key: string, uptoWeek: string): number
+  upsertWorklogComment(c: WorklogCommentRecord): { inserted: boolean }
+  hasWorklogComment(id: number): boolean
+  listWorklogComments(filters: WorklogFilters): WorklogCommentRecord[]
+  countWorklogComments(filters?: WorklogFilters): number
   rawDb(): Database.Database
   close(): void
 }
@@ -81,6 +110,25 @@ CREATE TABLE IF NOT EXISTS patterns (
 );
 CREATE INDEX IF NOT EXISTS idx_patterns_week ON patterns(week_iso);
 CREATE INDEX IF NOT EXISTS idx_patterns_key ON patterns(pattern_key);
+
+CREATE TABLE IF NOT EXISTS worklog_comments (
+  id              INTEGER PRIMARY KEY,
+  discussion_id   INTEGER NOT NULL,
+  created_at      TEXT NOT NULL,
+  staff_id        INTEGER,
+  title           TEXT,
+  task_ref        TEXT,
+  duration_sec    INTEGER,
+  work_items      TEXT NOT NULL,
+  files_modified  TEXT NOT NULL,
+  problems_json   TEXT NOT NULL,
+  patterns_text   TEXT NOT NULL,
+  actions_json    TEXT NOT NULL,
+  raw_html        TEXT NOT NULL,
+  imported_at     TEXT NOT NULL
+);
+CREATE INDEX IF NOT EXISTS idx_wc_discussion ON worklog_comments(discussion_id, created_at DESC);
+CREATE INDEX IF NOT EXISTS idx_wc_task ON worklog_comments(task_ref);
 `
 
 function rowToMeta(row: Record<string, unknown>): SessionMeta {
@@ -272,6 +320,89 @@ export function openSessionsDb(dbPath: string): SessionsDb {
       }
       return count
     },
+    upsertWorklogComment(c: WorklogCommentRecord): { inserted: boolean } {
+      const existing = db.prepare('SELECT 1 FROM worklog_comments WHERE id = ?').get(c.id)
+      const inserted = !existing
+      db.prepare(`
+        INSERT INTO worklog_comments (id, discussion_id, created_at, staff_id, title, task_ref,
+          duration_sec, work_items, files_modified, problems_json, patterns_text, actions_json,
+          raw_html, imported_at)
+        VALUES (@id, @discussion_id, @created_at, @staff_id, @title, @task_ref,
+          @duration_sec, @work_items, @files_modified, @problems_json, @patterns_text, @actions_json,
+          @raw_html, @imported_at)
+        ON CONFLICT(id) DO UPDATE SET
+          discussion_id = excluded.discussion_id,
+          created_at = excluded.created_at,
+          staff_id = excluded.staff_id,
+          title = excluded.title,
+          task_ref = excluded.task_ref,
+          duration_sec = excluded.duration_sec,
+          work_items = excluded.work_items,
+          files_modified = excluded.files_modified,
+          problems_json = excluded.problems_json,
+          patterns_text = excluded.patterns_text,
+          actions_json = excluded.actions_json,
+          raw_html = excluded.raw_html,
+          imported_at = excluded.imported_at
+      `).run({
+        id: c.id,
+        discussion_id: c.discussion_id,
+        created_at: c.created_at,
+        staff_id: c.staff_id,
+        title: c.title,
+        task_ref: c.task_ref,
+        duration_sec: c.duration_sec,
+        work_items: JSON.stringify(c.work_items),
+        files_modified: JSON.stringify(c.files_modified),
+        problems_json: JSON.stringify(c.problems),
+        patterns_text: JSON.stringify(c.patterns_text),
+        actions_json: JSON.stringify(c.actions),
+        raw_html: c.raw_html,
+        imported_at: c.imported_at,
+      })
+      return { inserted }
+    },
+    hasWorklogComment(id: number): boolean {
+      return !!db.prepare('SELECT 1 FROM worklog_comments WHERE id = ?').get(id)
+    },
+    listWorklogComments(filters: WorklogFilters): WorklogCommentRecord[] {
+      const parts: string[] = []
+      const params: Record<string, unknown> = {}
+      if (filters.discussion_id) { parts.push('discussion_id = @discussion_id'); params.discussion_id = filters.discussion_id }
+      if (filters.task_ref) { parts.push('task_ref = @task_ref'); params.task_ref = filters.task_ref }
+      if (filters.sinceIso) { parts.push('created_at >= @since'); params.since = filters.sinceIso }
+      const where = parts.length ? 'WHERE ' + parts.join(' AND ') : ''
+      const limit = filters.limit ?? 1000
+      const offset = filters.offset ?? 0
+      const rows = db.prepare(`SELECT * FROM worklog_comments ${where} ORDER BY created_at DESC LIMIT @limit OFFSET @offset`)
+        .all({ ...params, limit, offset }) as Record<string, unknown>[]
+      return rows.map((r) => ({
+        id: r.id as number,
+        discussion_id: r.discussion_id as number,
+        created_at: r.created_at as string,
+        staff_id: (r.staff_id as number | null) ?? null,
+        title: (r.title as string | null) ?? null,
+        task_ref: (r.task_ref as string | null) ?? null,
+        duration_sec: (r.duration_sec as number | null) ?? null,
+        work_items: JSON.parse(r.work_items as string),
+        files_modified: JSON.parse(r.files_modified as string),
+        problems: JSON.parse(r.problems_json as string),
+        patterns_text: JSON.parse(r.patterns_text as string),
+        actions: JSON.parse(r.actions_json as string),
+        raw_html: r.raw_html as string,
+        imported_at: r.imported_at as string,
+      }))
+    },
+    countWorklogComments(filters?: WorklogFilters): number {
+      const parts: string[] = []
+      const params: Record<string, unknown> = {}
+      if (filters?.discussion_id) { parts.push('discussion_id = @discussion_id'); params.discussion_id = filters.discussion_id }
+      if (filters?.task_ref) { parts.push('task_ref = @task_ref'); params.task_ref = filters.task_ref }
+      if (filters?.sinceIso) { parts.push('created_at >= @since'); params.since = filters.sinceIso }
+      const where = parts.length ? 'WHERE ' + parts.join(' AND ') : ''
+      const row = db.prepare(`SELECT COUNT(*) as c FROM worklog_comments ${where}`).get(params) as { c: number }
+      return row.c
+    },
     rawDb(): Database.Database {
       return db
     },