new features: translator api plus caching server-side, durable redis storage

new rule: NEVER LOG IN HIGH-LEVEL FUNCTIONS!!!!

app/ ├── core/ # shared game logic - knows nothing about HTTP │ ├── combat.py │ ├── dungeon.py │ ├── progression.py │ └── translation.py │ ├── api/ # FastAPI - thin layer, calls core │ └── routes/ │ ├── dungeon.py # calls core.dungeon │ └── combat.py # calls core.combat │ └── workers/ # also calls core, not the API └── handlers/ └── loot.py # calls core.progression

current progress: game WIP. translator finished, all tests pass.

also, game MANAGES DATA SOLO. API has nothing to do with it. SoC must be respected!

i decided to abstract away slots and other metadata - they shouldn't figure in phrases.py

TRANSLATION LAYER: at launch for every user, depending on language, the phrases are either being all translated or not.

Translation caching behavior:

- backend/services/translator/translation_service.py.translate does not write to Redis; it only resolves text.

- backend/services/cache_warm.py.warm_cache generates precomputed strings for non-dynamic phrase values and stores cache

entries as translation:{lang}:{hash(text)}.

- Dynamic phrases (containing {} markers) are excluded from warm cache via get_cacheable.

Translator API + Redis caching

/translate endpoint in backend/api/main.py:

• Query params: text, to_language (default en)
• Uses backend.services.translator.translation_service.translate
• Reads and writes cache in Redis using translation:{lang}:{hash(text)}
• TTL controlled by backend/services/translator/config.py (cache_time)

New test: tests/services/test_translator_api.py