app/routers/chat.py

@@ -2,7 +2,7 @@
 app/routers/chat.py — RAG Endpunkt (WP-05)
 
 Version:
-    0.1.1 (Debug-Logging enabled)
+    0.2.0 (Final Clean Version)
 """
 
 from fastapi import APIRouter, HTTPException, Depends
@@ -13,7 +13,6 @@ import logging
 
 from app.models.dto import ChatRequest, ChatResponse, QueryRequest, QueryHit
 from app.services.llm_service import LLMService
-# Annahme: Der Retriever aus WP-04 liegt hier.
 from app.core.retriever import Retriever 
 
 router = APIRouter()
@@ -32,20 +31,16 @@ def _build_context_from_hits(hits: List[QueryHit]) -> str:
     context_parts = []
     for i, hit in enumerate(hits, 1):
         source = hit.source or {}
-        # Wir probieren alle möglichen Felder, in denen Text stecken könnte
+        # Robuster Zugriff auf Content
         content = (
             source.get("text") or 
             source.get("content") or 
             source.get("page_content") or 
-            source.get("chunk_text") or # Oft verwendet
-            "[[LEERER INHALT - PAYLOAD PRÜFEN]]"
+            source.get("chunk_text") or 
+            "[Kein Textinhalt verfügbar]"
         )
-        
         title = hit.note_id or "Unknown Note"
         
-        # Debug Log für jeden Hit
-        logger.info(f"Building Context [{i}]: ID={hit.node_id} Content-Length={len(str(content))}")
-        
         entry = (
             f"SOURCE [{i}]: {title} (Score: {hit.total_score:.2f})\n"
             f"CONTENT: {content}\n"
@@ -63,7 +58,8 @@ async def chat_endpoint(
     start_time = time.time()
     query_id = str(uuid.uuid4())
     
-    logger.info(f"Chat request [{query_id}]: {request.message}")
+    # Minimales Logging für Traceability
+    logger.info(f"Chat request [{query_id}]: {request.message[:50]}...")
 
     try:
         # 1. Retrieval
@@ -79,15 +75,13 @@ async def chat_endpoint(
 
         # 2. Kontext bauen
         if not hits:
-            logger.info(f"[{query_id}] No hits found for context.")
+            logger.info(f"[{query_id}] No hits found.")
             context_str = "Keine relevanten Notizen gefunden."
         else:
             context_str = _build_context_from_hits(hits)
-            
-        # WICHTIG: Wir loggen den ersten Teil des Kontextes, um zu sehen, was das LLM bekommt
-        logger.info(f"--- LLM CONTEXT PREVIEW ---\n{context_str[:500]}\n--- END PREVIEW ---")
 
         # 3. LLM Generation
+        logger.info(f"[{query_id}] Sending to LLM ({len(hits)} context chunks)...")
         answer_text = await llm.generate_rag_response(
             query=request.message,
             context_str=context_str
@@ -95,6 +89,7 @@ async def chat_endpoint(
 
         # 4. Response
         duration_ms = int((time.time() - start_time) * 1000)
+        logger.info(f"[{query_id}] Completed in {duration_ms}ms")
         
         return ChatResponse(
             query_id=retrieve_result.query_id,