i want to create a intern company knowledge database. The KI is fed with internal knowledge
information (e.g., organizational handbook, all contents of homepage, order forms, internal briefing and training documents for campaigns).
And is then available to the employee as an “omniscient colleague” who can be asked.
ChatGPT itsself is very general in its answer.
So how to to this ?
Use embeddings to find data relevant to the query using cosine similarly, then supply the context and question to the model for completion.
Okay, i seen some source code with langchain. Seems really good.
Can i store this embeddings permanent, so that they are persistant in a database? In the example they used FAISS.
is using embedding or langchain will give same exact answer from knowledge base
or it will aslo add something creative to the answer