Does this project use any special trick to retrieve best possible documents for answering query/question from our docs ?

[Greatz08]

I have self-hosted openwebui and i did selfhost milvus vector db and connect both to vectorize my .md files notes related to cybersecurity, so that it can retrieve best possible documents to answer my query, but in my latest experience sometimes it does fetch right document to answer and sometimes it doesn't. The more similar data notes i have the more problematic it makes for it to fetch, so i was thinking like if this project use any special technique behind the scene to fetch most relevant document to answer query or it is based on same principles like that of openwebui where we connect with vector db and upload content and let it chunk 1000 character and expect it to fetch most relevant ones.