[PGVector] As a user I want to create and segregate embeddings results into different tables based on different type of search scope and enable FGAC

[subhsaha]

Checked

• I searched existing ideas and did not find a similar one
• I added a very descriptive title
• I've clearly described the feature request and motivation for it

Feature request

PGVector currently creates two new tables, langchain_pg_embedding and langchain_pg_collection, if they do not already exist. If the table exists it loads the data into existing tables. However, if I want to create and segregate embedding results into different tables based on different search scopes,it is not available. For Example -

current

db = PGVector.from_documents(
    documents= docs,
    embedding = embeddings,
    collection_name= "collection_name_a",
      distance_strategy = DistanceStrategy.COSINE,
    connection_string=CONNECTION_STRING,
    use_jsonb=True

    )

Proposed

db = PGVector.from_documents(
    documents= docs,
    embedding = embeddings,
    collection_name= "collection_name_b",
    table_name= "table_to_store_embedding", --if table dose not exist  new table will be created
    distance_strategy = DistanceStrategy.COSINE,
    connection_string=CONNECTION_STRING,
  
    use_jsonb=True

    )

Q) What happens to langchain_pg_collection ?
A: langchain_pg_collection stays same with additional column for 'table_name'

Motivation

Providing option to segregate enables the implementation of Fine Grain Access Control .Here is a discussion - #6045

Proposal (If applicable)

Proposed

db = PGVector.from_documents(
    documents= docs,
    embedding = embeddings,
    collection_name= "collection_name_b",
    table_name= "table_to_store_embedding", --if table dose not exist  new table will be created
    distance_strategy = DistanceStrategy.COSINE,
    connection_string=CONNECTION_STRING,
  
    use_jsonb=True

    )

Q) What happens to langchain_pg_collection ?
A: langchain_pg_collection stays same with additional column for 'table_name'

[stinkymonkeyph]

Where are we with this?