feat: replace manual web scraping with an agentic tool-selection engine

feat: implement Redis-first refresh token storage with Supabase fallback

feat: migrate RAG chatbot model from Groq Llama 3.1 to Ministral 8b

feat: implement LLM fallback for quiz generation and enhance OAuth error handling

feat: optimize auth concurrency and enable full account deletion

feat: implement in-memory email rate limiting and adjust RAG search parameters

feat: fix LLM model names, increase RAG retrieval and memory limits

feat: improve chatbot search quality, implement refusal filtering.

refactor: optimize RAG prompt instructions, update web search tool configuration, and streamline search logic.

feat: add religious, political and NSFW topics in prohibited content criteria in quiz and summary generators

refactor: update system prompts safety guidelines

refactor: update content limits for quiz and summary generation modules

refactor: changing to gemini model, modularize constants and schemas and update user daily generation limit