From 7672bbcad69e6ed2043574a9317105ddb9cde4c5 Mon Sep 17 00:00:00 2001
From: claude-dev <claude-dev@aegis-sight.de>
Date: Wed, 4 Mar 2026 23:28:37 +0100
Subject: [PATCH] =?UTF-8?q?Domain-Balance=20vor=20Haiku=20entfernt=20?=
 =?UTF-8?q?=E2=80=94=20Feeds=20sollen=20alle=20verf=C3=BCgbar=20bleiben?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Die Feed-Vorfilterung (max 3 pro Domain) vor der Haiku-Selektion war
falsch: Alle thematischen Feeds (z.B. Guardian World, Politics, Middle East)
sollen Haiku zur Auswahl stehen. Die Quellenvielfalt wird stattdessen
durch den Prompt (QUELLENVIELFALT-Regel) und den Artikel-Cap nach
dem RSS-Fetch (max 10 Artikel/Domain) sichergestellt.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
---
 src/agents/orchestrator.py | 33 +--------------------------------
 1 file changed, 1 insertion(+), 32 deletions(-)

diff --git a/src/agents/orchestrator.py b/src/agents/orchestrator.py
index 3cc96bf..501b1d5 100644
--- a/src/agents/orchestrator.py
+++ b/src/agents/orchestrator.py
@@ -4,9 +4,8 @@ import json
 import logging
 import re
 from datetime import datetime, timezone
-from config import TIMEZONE, MAX_FEEDS_PER_DOMAIN
+from config import TIMEZONE
 from typing import Optional
-from collections import defaultdict
 from urllib.parse import urlparse, urlunparse
 
 from agents.claude_client import UsageAccumulator
@@ -163,14 +162,6 @@ 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(),),
@@ -587,28 +578,6 @@ 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(