Notes

in this chat popup as discussed previously shows on sermon producer, we can create knowledge base with data of following sources which we would probably have in the first place,
1. A simple documentation for each miniteries,
2. A simple documentation for each church inside each ministeries,
3. A intro documentation for pastor or sermons of each chirch
4. then we have video's of the respective chirch attended by respective paster or sermon where we have 2 kind of information about the video: -
4.1. Detailed description about the video with full transcriptions and their echo /post details,
4.2. Newsletter content of the respective video

Based on these knowledge base where we can have shared and connected knowledge data, we can create proper RAG pipeline where we can follow following steps: -
Loading in MD type
Semmantic Chunking
High Dimention multimodel embedding
peroper vectorstore storage with indexing technique like Approximate nearest naighbour in it
Use BM25 + Embedding based Hybrid retriever
Reranker
Query Builder
LLM
Chain-Runnables
LLM as a gudge with RAGAS performance parameter checkup
Guardrail output Parser in JSON

this chatbot will be interactive and live sync with the knowledge data amongs all the sermon platform.