Feat: add support for initializing vecs client with custom schema

[majamil16]

What kind of change does this PR introduce?

Feature: add support for connecting to a user-specified database schema (instead of hard-coding vecs).

This was accomplished by adding the schema argument to the Client class and replacing all hard-coded references to schema vecs.

Also added schema parameter to create_client method. For backwards compatibility, it still defaults to vecs.

What is the current behavior?

Currently, the client defaults to schema vecs. Users may want to place vecs collections under a schema according to their own naming conventions.

Addresses #62

Please link any relevant issues here.

What is the new behavior?

Users can now instantiate a Client instance with a custom schema. For example:

DB_CONNECTION = "postgresql://<user>:<password>@<host>:<port>/<db_name>"

# create vector store client
vx = vecs.create_client(DB_CONNECTION, schema="my_schema")

Additional context

Add any other context or screenshots.

Test coverage attached

[olirice]

nice job on the PR

did you see this comment here:
#62 (comment)

if we add a top level parameter for schema in the Client it'll be much more difficult to support defining a schema on the Collection when that inevitably gets requested.

Would you be up for refactoring to make this a per Collection keyword only argument?

Note that it'd be a bit more involved since the Metadata instance can't be bound to a single schema in that case

[majamil16]

@olirice thanks for the feedback! sure, I'm definitely up to implement that change.

[majamil16]

hey @olirice, made some updates. let me know your thoughts, thanks!

[olirice]

looks great. if we can address those last few concerns this will merge

[yusofy]

I'm looking forward to this PR being merged!

[majamil16]

hey, thanks @olirice for the review! will finish this up this week

[olirice]

 and optional schema. Defaults to `vecs` schema.

can be removed

[olirice]

what was the rationale for this change?

[olirice]

the best way to test that escaping is done correctly in all places is to use a crazy schema name.

I tested basic operations with the schema name "esCape Me!" and the test suite fails.

To reproduce that, try:

foo: Collection = client.get_or_create_collection(name="foo", schema="esCape Me!", dimension=5)

and you'll get

sqlalchemy.exc.ProgrammingError: (psycopg2.errors.InvalidSchemaName) schema ""esCape Me!"" does not exist

[ghost]

Hi Guys !

Any news about it ?

[faridkar]

Hi - any news? For me personally, this change would greatly improve my experience using vecs!

[diksipav]

Hey @olirice I can take and finish this if @majamil16 is not going to?

[olirice]

thanks! but lets hold off until we land the other PR you have open