Quellenvielfalt sicherstellen: Domain-Cap + Balance + Discovery-Verbesserungen

- config.py: MAX_FEEDS_PER_DOMAIN=3, MAX_ARTICLES_PER_DOMAIN_RSS=10
- rss_parser.py: _apply_domain_cap() begrenzt Artikel pro Domain nach RSS-Fetch
- orchestrator.py: Domain-Balance vor Feed-Selektion (max 3 Feeds/Domain),
  Domain-Cap in Background-Discovery
- source_rules.py: article_count in get_feeds_with_metadata(), Content-Hash
  in _validate_feed() für Duplikat-Erkennung bei Discovery
- researcher.py: QUELLENVIELFALT-Regel im Haiku Feed-Selektions-Prompt
- DB: 52 WordPress-Redirect-Duplikate deaktiviert (netzpolitik.org, bashinho.de)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

src/agents/orchestrator.py

@@ -4,8 +4,9 @@ import json
 import logging
 import re
 from datetime import datetime, timezone
-from config import TIMEZONE
+from config import TIMEZONE, MAX_FEEDS_PER_DOMAIN
 from typing import Optional
+from collections import defaultdict
 from urllib.parse import urlparse, urlunparse
 
 from agents.claude_client import UsageAccumulator
@@ -162,6 +163,14 @@ async def _background_discover_sources(articles: list[dict]):
         # 3. Gegen DB prüfen — welche Domains existieren schon?
         new_count = 0
         for domain, url, category in domains_to_check:
+            cursor = await db.execute(
+                "SELECT id FROM sources WHERE LOWER(domain) = ? AND source_type = 'rss_feed' AND status = 'active'",
+                (domain.lower(),),
+            )
+            existing_feeds = await cursor.fetchall()
+            if len(existing_feeds) >= MAX_FEEDS_PER_DOMAIN:
+                continue  # Domain hat bereits genug aktive Feeds
+
             cursor = await db.execute(
                 "SELECT id FROM sources WHERE LOWER(domain) = ?",
                 (domain.lower(),),
@@ -578,6 +587,28 @@ class AgentOrchestrator:
                 from source_rules import get_feeds_with_metadata
                 all_feeds = await get_feeds_with_metadata(tenant_id=tenant_id)
 
+                # Domain-Balance: Max. MAX_FEEDS_PER_DOMAIN Feeds pro Domain
+                feeds_by_domain: dict[str, list[dict]] = defaultdict(list)
+                for feed in all_feeds:
+                    feeds_by_domain[feed.get("domain", "")].append(feed)
+
+                balanced_feeds = []
+                for domain, domain_feeds in feeds_by_domain.items():
+                    if len(domain_feeds) > MAX_FEEDS_PER_DOMAIN:
+                        # Nach article_count sortieren, meistgenutzte behalten
+                        domain_feeds.sort(key=lambda f: f.get("article_count", 0), reverse=True)
+                        kept = domain_feeds[:MAX_FEEDS_PER_DOMAIN]
+                        logger.info(
+                            f"Domain-Balance: {domain} von {len(domain_feeds)} auf {MAX_FEEDS_PER_DOMAIN} Feeds begrenzt"
+                        )
+                        balanced_feeds.extend(kept)
+                    else:
+                        balanced_feeds.extend(domain_feeds)
+
+                if len(balanced_feeds) < len(all_feeds):
+                    logger.info(f"Domain-Balance gesamt: {len(all_feeds)} → {len(balanced_feeds)} Feeds")
+                all_feeds = balanced_feeds
+
                 feed_usage = None
                 if len(all_feeds) > 20:
                     selected_feeds, feed_usage = await rss_researcher.select_relevant_feeds(