Update samples: Import PebbloRetrievalQA from langchain_community #364
Conversation
Raj725
changed the title
|
NOTE: |
| @@ -80,6 +80,9 @@ def init_retrieval_chain(self): | |||
| """ | |||
| return PebbloRetrievalQA.from_chain_type( | |||
| llm=self.llm, | |||
| app_name=self.app_name, | |||
There was a problem hiding this comment.
With safe retrieval details on Local UI, we now have two different types of apps - loader and retriever. The discover call from PebbloSafeLoader would create loader type app where as discover call from PebbloRetrievalQA would create retriever type app. If we use same app name at both places, it will show only one app which is retriever, because discover from PebbloRetrievalQA was made later.
@srics This is exact corner case that we were talking while deciding approach for loader and retriever type applications.
TLDR - Either we should use separate names as app_name parameter to PebbloRetrievalQA.from_chain_type() and PebbloSafeLoader or we should have two separate applications one of loader type and other of retrieval type with separate names.
There was a problem hiding this comment.
We can adjust the naming in our sample apps, but we cannot prevent users from naming them the same.
Logically, it's one app that loads data from documents and allows users to query them.
There was a problem hiding this comment.
@shreyas-damle @Raj725 one data ingestion app could potentially serve multiple retrieval/chatbot apps. It is reasonable to let the app owners provide unique names across pebblo. We can relax this requirement based on user feedback and/or when we switch over to using sqlite.
There was a problem hiding this comment.
@shreyas-damle Updated
Raj725
changed the title
username->user_id