Merge pull request #53 from shin902/feat/rss-feed-polling

feat: RSSフィードの自動配信機能を追加

Author: shin902

.claude/skills/setup-rss/SKILL.md

@@ -0,0 +1,87 @@
+---
+name: setup-rss
+description: Configure RSS feed polling for NanoClaw channels. Adds rss.channels section to nanoclaw.yaml with per-channel feed URLs, and restarts the service. Triggers on "setup rss", "add rss", "rss feed", or "rss 配信".
+---
+
+# RSS Feed Setup
+
+Add RSS feed polling to NanoClaw so new articles are automatically delivered to configured channels every 15 minutes.
+
+## 1. Check Current Configuration
+
+Read `nanoclaw.yaml` to see if an `rss` section already exists:
+
+```bash
+cat nanoclaw.yaml
+```
+
+If the file doesn't exist, create it with just the `rss` section. If it already has a `providers` section, add the `rss` section alongside it.
+
+## 2. Ask User for Configuration
+
+Use `AskUserQuestion` to ask:
+
+1. **Which channel ID(s)** should receive RSS updates? (e.g., `dc:1234567890`)
+2. **Which feed URLs?** — one or more RSS/Atom feed URLs to monitor.
+
+## 3. Add RSS Section to nanoclaw.yaml
+
+Add or update the `rss` section in `nanoclaw.yaml`. The structure is:
+
+```yaml
+rss:
+  channels:
+    - jid: "dc:1234567890"
+      feeds:
+        - url: "https://example.com/feed.xml"
+          name: "Example Blog"
+        - url: "https://another.example/rss"
+```
+
+- `jid`: The channel JID (must match a registered group in NanoClaw)
+- `feeds[].url`: RSS/Atom feed URL (required)
+- `feeds[].name`: Display name for the feed (optional, used in messages)
+
+Use the Edit tool to insert the `rss` section. If `nanoclaw.yaml` already has content, add `rss` at the same level as `providers`.
+
+## 4. Verify Channel Registration
+
+Ensure the JIDs specified in the RSS config correspond to registered groups:
+
+```bash
+sqlite3 store/messages.db "SELECT jid, name, folder FROM registered_groups"
+```
+
+If a JID is not registered, RSS items for that channel will be skipped with a warning log. Register the channel first using the NanoClaw interface.
+
+## 5. Restart NanoClaw
+
+```bash
+# macOS
+launchctl kickstart -k gui/$(id -u)/com.nanoclaw
+
+# Linux
+systemctl --user restart nanoclaw
+
+# Or manual restart
+npm run dev
+```
+
+The RSS poller starts automatically on boot and checks for new articles every 15 minutes.
+
+## 6. Verify It Works
+
+Check the logs for RSS poller startup:
+
+```bash
+tail -f logs/nanoclaw.log | grep -i rss
+```
+
+You should see: `RSS poller started (15-minute interval)` and `RSS config loaded from nanoclaw.yaml`.
+
+## Notes
+
+- RSS items are tracked in the `rss_seen_items` SQLite table — restarts do not cause duplicate deliveries.
+- The `guid` element is used as the unique item identifier, falling back to the article URL.
+- Feed fetch failures are logged as warnings and do not crash the poller.
+- To add or remove feeds, edit `nanoclaw.yaml` and restart. Config is re-read on each poll cycle.

package-lock.json

@@ -19,8 +19,9 @@
   },
   "dependencies": {
     "better-sqlite3": "^11.8.1",
-    "discord.js": "^14.18.0",
     "cron-parser": "^5.5.0",
+    "discord.js": "^14.18.0",
+    "fast-xml-parser": "^5.7.2",
     "pino": "^9.6.0",
     "pino-pretty": "^13.0.0",
     "yaml": "^2.8.2",

package.json

@@ -274,6 +274,15 @@ function createSchema(database: Database.Database): void {
     );
     CREATE INDEX IF NOT EXISTS idx_spawned_threads_created_at
       ON spawned_threads(created_at);
+
+    CREATE TABLE IF NOT EXISTS rss_seen_items (
+      feed_url TEXT NOT NULL,
+      item_id TEXT NOT NULL,
+      seen_at TEXT NOT NULL,
+      PRIMARY KEY (feed_url, item_id)
+    );
+    CREATE INDEX IF NOT EXISTS idx_rss_seen_items_seen_at
+      ON rss_seen_items(seen_at);
   `);
 
   // sessions テーブルのスキーマを group_folder → group_jid に移行。
@@ -470,6 +479,25 @@ function startSpawnedThreadsCleanupTimer(): void {
   spawnedThreadsCleanupTimer.unref();
 }
 
+const RSS_SEEN_ITEMS_RETENTION_DAYS = 90;
+const RSS_CLEANUP_INTERVAL_MS = 60 * 60 * 1000;
+let rssCleanupTimer: ReturnType<typeof setInterval> | null = null;
+
+function startRssCleanupTimer(): void {
+  if (rssCleanupTimer) {
+    return;
+  }
+
+  rssCleanupTimer = setInterval(() => {
+    try {
+      cleanupRssSeenItems();
+    } catch (error) {
+      logger.warn({ err: error }, 'Failed to clean up rss_seen_items');
+    }
+  }, RSS_CLEANUP_INTERVAL_MS);
+  rssCleanupTimer.unref();
+}
+
 export function initDatabase(): void {
   const dbPath = path.join(STORE_DIR, 'messages.db');
   fs.mkdirSync(path.dirname(dbPath), { recursive: true });
@@ -481,6 +509,8 @@ export function initDatabase(): void {
   migrateJsonState();
   cleanupSpawnedThreads();
   startSpawnedThreadsCleanupTimer();
+  cleanupRssSeenItems();
+  startRssCleanupTimer();
 }
 
 /** @internal - テスト用のみ。新規のインメモリデータベースを作成します。 */
@@ -490,6 +520,11 @@ export function _initTestDatabase(): void {
     spawnedThreadsCleanupTimer = null;
   }
 
+  if (rssCleanupTimer) {
+    clearInterval(rssCleanupTimer);
+    rssCleanupTimer = null;
+  }
+
   db = new Database(':memory:');
   createSchema(db);
 }
@@ -1206,6 +1241,70 @@ export function cleanupSpawnedThreads(
   return result.changes;
 }
 
+// --- RSS 既読管理 ---
+
+export function hasSeenItem(feedUrl: string, itemId: string): boolean {
+  const row = db
+    .prepare('SELECT 1 FROM rss_seen_items WHERE feed_url = ? AND item_id = ?')
+    .get(feedUrl, itemId);
+  return row !== undefined;
+}
+
+const SQLITE_VARIABLE_LIMIT = 900;
+
+export function getSeenItemIds(
+  feedUrl: string,
+  itemIds: string[],
+): Set<string> {
+  if (itemIds.length === 0) return new Set();
+  const seen = new Set<string>();
+  for (let i = 0; i < itemIds.length; i += SQLITE_VARIABLE_LIMIT) {
+    const chunk = itemIds.slice(i, i + SQLITE_VARIABLE_LIMIT);
+    const placeholders = chunk.map(() => '?').join(',');
+    const rows = db
+      .prepare(
+        `SELECT item_id FROM rss_seen_items WHERE feed_url = ? AND item_id IN (${placeholders})`,
+      )
+      .all(feedUrl, ...chunk) as { item_id: string }[];
+    for (const r of rows) seen.add(r.item_id);
+  }
+  return seen;
+}
+
+export function markItemSeen(feedUrl: string, itemId: string): void {
+  db.prepare(
+    'INSERT OR IGNORE INTO rss_seen_items (feed_url, item_id, seen_at) VALUES (?, ?, ?)',
+  ).run(feedUrl, itemId, new Date().toISOString());
+}
+
+/** @internal - テスト用のみ。seen_at を強制上書きする。 */
+export function _forceRssSeenItemTimestamp(
+  feedUrl: string,
+  itemId: string,
+  seenAt: string,
+): void {
+  db.prepare(
+    `UPDATE rss_seen_items SET seen_at = ? WHERE feed_url = ? AND item_id = ?`,
+  ).run(seenAt, feedUrl, itemId);
+}
+
+export function cleanupRssSeenItems(
+  now: Date = new Date(),
+  retentionDays: number = RSS_SEEN_ITEMS_RETENTION_DAYS,
+): number {
+  const cutoff = new Date(now.getTime() - retentionDays * 24 * 60 * 60 * 1000);
+  const result = db
+    .prepare(`DELETE FROM rss_seen_items WHERE seen_at < ?`)
+    .run(cutoff.toISOString());
+  if (result.changes > 0) {
+    logger.info(
+      { deletedRows: result.changes, retentionDays },
+      'Cleaned up stale rss_seen_items rows',
+    );
+  }
+  return result.changes;
+}
+
 // --- JSON マイグレーション ---
 
 function migrateJsonState(): void {

src/db.ts

@@ -60,6 +60,7 @@ import {
   loadSenderAllowlist,
   shouldDropMessage,
 } from './sender-allowlist.js';
+import { startRssPoller } from './rss-poller.js';
 import { startSchedulerLoop } from './task-scheduler.js';
 import { hasPrivilege, resolveGroupType } from './group-type.js';
 import {
@@ -1013,6 +1014,23 @@ async function main(): Promise<void> {
       if (text) await channel.sendMessage(jid, text);
     },
   });
+  startRssPoller({
+    sendMessage: async (jid, text) => {
+      // pollOnce also checks this before fetching feeds (avoids network cost).
+      // This guard is a safety net in case sendMessage is reached via other paths.
+      if (!(jid in registeredGroups)) {
+        logger.warn({ jid }, 'RSS: jid not in registered groups, skipping');
+        return;
+      }
+      const channel = findChannel(channels, jid);
+      if (!channel) {
+        logger.warn({ jid }, 'RSS: no channel owns JID, skipping');
+        return;
+      }
+      await channel.sendMessage(jid, formatOutbound(text));
+    },
+    registeredGroups: () => registeredGroups,
+  });
   startIpcWatcher({
     sendMessage: (jid, text) => {
       const channel = findChannel(channels, jid);