harden information getter services and optimize user msg

config/retriex/prompt.yaml

@@ -47,6 +47,7 @@ parameters:
       conversation_context_label: CONVERSATION CONTEXT (contextual only)
       shop_search_query_label: SHOP SEARCH QUERY
       output_priority_label: OUTPUT PRIORITY
+      fallback_escalation_label: FALLBACK AND ESCALATION RULES
       response_format_label: RESPONSE FORMAT RULES
       language_rules_label: LANGUAGE RULES
       fact_grounding_rules_label: FACT GROUNDING RULES
@@ -72,6 +73,36 @@ parameters:
       - '- If one source chunk contains both the best matching value and nearby comparison values, use the nearby values only as context and do not include them unless the user asks for comparison or alternatives.'
       - '- For lowest/highest/minimum/maximum questions, answer only the requested extreme value and the product/device explicitly connected to it.'
       - '- Do not add runner-up products, second-lowest values, adjacent ranges, broader tables, or explanatory comparisons unless explicitly requested.'
+    fallback_escalation:
+      state_line_template: '- Internal confidence state: {state}.'
+      base_rules:
+      - '- Prefer transparent uncertainty over a confident but unsupported answer.'
+      - '- Never present missing or weak evidence as proof that a product, value, accessory, or suitability does not exist.'
+      - '- A negative answer is allowed only when the provided sources explicitly support that negative finding for the asked scope.'
+      - '- If several products, parameters, or accessories could match, ask one focused clarification question instead of guessing.'
+      - '- For risky or binding product selection, state that sales or support should verify the application before a final selection.'
+      without_shop_check_rules:
+      - '- If the question is product-related and no live shop check was performed in this run, do not make a portfolio-wide negative statement such as "there is no product".'
+      - '- Phrase missing evidence narrowly, for example: "Im RAG-Wissen finde ich dazu keine belastbare Information."'
+      - '- If useful, say that a shop search can be used to look for matching products, but do not claim shop results were checked unless they are present in the prompt.'
+      states:
+        sicher_beantwortbar:
+        - '- The retrieved factual knowledge or user-provided URL content is sufficient for the core answer. Answer directly, but do not exceed the provided facts.'
+        wahrscheinlich_beantwortbar:
+        - '- Retrieved knowledge and shop data are both available. Use retrieved knowledge for technical suitability and shop data for current commercial details.'
+        - '- If the two source types do not clearly refer to the same product identity, separate the technical answer from commercial shop hits.'
+        nur_shop_treffer_kein_belastbares_fachwissen:
+        - '- Start the answer by making the fallback clear: "Aus den Shopdaten ergeben sich folgende Treffer; technische Eignung bitte prüfen."'
+        - '- Do not present shop-only matches as verified technical suitability unless the shop text explicitly states that suitability.'
+        - '- Do not say that RAG knowledge confirms the result. Say that no belastbares RAG-Fachwissen was available for this selection.'
+        keine_belastbaren_daten:
+        - '- State that no reliable information was found in the provided RAG knowledge, URL content, or shop results.'
+        - '- Do not answer with "gibt es nicht". Use narrow wording such as "Ich finde dazu keine belastbaren Daten in den vorliegenden Quellen."'
+        - '- Ask one focused clarification question if a parameter, product family, accessory type, or application context would make the search answerable.'
+        shopdaten_nicht_verfuegbar:
+        - '- State that live shop data could not be loaded and answer only from retrieved knowledge or URL content if available.'
+        - '- Do not draw negative conclusions about current product availability, price, or shop portfolio while the shop is unavailable.'
+
     response_format:
       base_rules:
       - '- Keep normal spacing between all words. Never fuse words together.'