MtoRagSystem/RETRIEX_VOCABULARY_FIX_README.md

RetrieX Vocabulary Centralization Fix

This patch centralizes the growing recognition word lists without changing their tuned content.

Main changes

• Added config/retriex/vocabulary.yaml.
• Added App\Config\DomainVocabularyConfig.
• Wired the vocabulary facade into:
  • ShopServiceConfig
  • NdjsonHybridRetrieverConfig
  • PromptBuilderConfig
  • CommerceQueryParserConfig
• Moved the active Shop and Retrieval vocabulary defaults out of commerce.yaml and retrieval.yaml into vocabulary.yaml.
• Kept all old per-service config keys as explicit overrides.
• Removed direct NdjsonHybridRetrieverConfig::... constant usage inside NdjsonHybridRetriever so effective config getters are used consistently.

Stability note

The vocabulary views preserve the previous order and content of the tuned lists. No new semantic terms were added to the critical retrieval and shop matching views.

Required regression baseline:

• Was ist der niedrigste Grenzwert für die Wasserhärte, welcher mit einem Testomaten überwacht werden kann?
  • expected: 0,02 °dH (Testomat 808)
• mit welchem indikator wird der wert gemessen
  • expected: Indikatortyp 300
• Store query with 0,02 must preserve the decimal value and must not turn it into 02.