[ENHANCEMENT]: [Performance] Add missing cache decorator to get_category_books #576

Current Behavior

Currently, the get_category_books() function in backend/ai_service.py handles the generation of relevant book lists based on specific shelf themes and vibes. However, other expensive LLM operations in this service (such as get_ai_recommendations or generate_chat_response), this function completely lacks a caching decorator.

As a result, every time a user refreshes the virtual library or navigates the landing page shelves, a raw, LLM network request is fired to generate the data from scratch.

in a nutshell,
The current behavior forces the backend to call the AI from scratch every single time a user loads or refreshes a book shelf. This slows down the app to a crawl and burns through your API limits on identical requests that could easily be saved and reused.

kindly consider assigning it to me under GSSOC'26 .

Suggested Improvement

• Apply Decorator-Based Caching: Leverage the existing caching ecosystem established in backend/cache_service.py to seamlessly wrap get_category_books().

• Category-Specific Cache Keys: Ensure the underlying cache key dynamically hashes both the category string and the requested count so that varying limits don't serve mismatched cached arrays

Benefits

• Drops response times for shelf fetching from several seconds (waiting on LLM generation) to milliseconds (reading from local cache memory/Redis).

• Drastically minimizes the number of duplicate API tokens consumed from upstream providers (Groq, Gemini, or OpenAI), preventing rapid quota exhaustion during multi-user testing.

• Prevents the frontend virtual library shelves from looking blank or displaying skeleton loaders repeatedly on minor page update

Possible Implementation

No response