This is what I am currently doing:

At a glance

The community member is setting up a PGVectorStore with various parameters, including the database, table name, and embedding dimension. The comments discuss that while new columns can be added manually to the table, they won't be used for querying unless the object class is modified. However, the metadata column can be used for filtering. A community member asks for an example or tutorial on using metadata filtering with Postgres hybrid search, and another community member provides a link to the relevant documentation.

Useful resources

AAnurag Agrawal

This is what I am currently doing:

pg_vector_store = PGVectorStore.from_params(
database=db_name,
host=url.host,
password=url.password,
port=url.port,
user=url.username,
table_name="nodes",
embed_dim=384, # bge-small-v1.5 embedding dimension
hybrid_search=True,
text_search_config="english",
)

7 comments

LLogan M

sadly no. I mean, you can add new columns manually to the table, but they wont be used when querying unless you modify the actual object class

LLogan M

Still thinking/working on the best kind of way (if possible) to provide what you ware wanting 🙂

LLogan M

You CAN query/filter on the metadata column though

LLogan M

if it's in the metadata

AAnurag Agrawal

Thanks Logan! Is there an example/tutorial for using metadata filtering with postgres hybrid search?

LLogan M

Every vectordb uses the same metadata-filtering concept. You can see an example here:

https://docs.llamaindex.ai/en/stable/optimizing/basic_strategies/basic_strategies.html#metadata-filters

AAnurag Agrawal

Thanks @Logan M!

Add a reply

Find answers from the community

This is what I am currently doing: