I like what you are doing! I suggest you create a proxy service that will route the traffic to the AI provider.

You can cache questions and answers heavily and use a B25 search with a vector embedded to retrieve the best results for you.

   RAG pipeline
   ├── BM25 + TF-IDF + RRF retrieval
   ├── cross-encoder reranking
   ├── knowledge-graph entity linking
   └── multi-angle intent detection
        │
        ▼
   LLM synthesis  (Claude / local models)