The lack of feedback during Collection.upsert and Collection.create_index is a bad DX.
It would be great to get some progress bars but I haven't been able to get them working properly in notebooks and shell environments.
If anyone has experience adding them this would be a great community contribution
upsert
|
for chunk in flu(vectors).chunk(chunk_size): |
but it can't always assume that
vectors is sized. Maybe a runtime check to see if
*vectors* has a known length and providing it to the progress bar if known would be best
create_index
|
stmt = postgresql.insert(clone_table).from_select( |
|
self.table.c, select(self.table) |
|
) |
|
stmt = stmt.on_conflict_do_nothing() |
|
sess.execute(stmt) |
this one is a little more involved as it'll have to introduce client side keyset pagination on the id primary key to get feedback to python
The lack of feedback during
Collection.upsertandCollection.create_indexis a bad DX.It would be great to get some progress bars but I haven't been able to get them working properly in notebooks and shell environments.
If anyone has experience adding them this would be a great community contribution
upsert
vecs/src/vecs/collection.py
Line 108 in 87ed2d3
but it can't always assume that
vectorsis sized. Maybe a runtime check to see if*vectors*has a known length and providing it to the progress bar if known would be bestcreate_index
vecs/src/vecs/collection.py
Lines 346 to 350 in 87ed2d3
this one is a little more involved as it'll have to introduce client side keyset pagination on the
idprimary key to get feedback to python