apps/docs/content/guides/ai/google-colab.mdx
<a className="w-64" href="https://colab.research.google.com/github/supabase/supabase/blob/master/examples/ai/vector_hello_world.ipynb"
</a>
Google Colab is a hosted Jupyter Notebook service. It provides free access to computing resources, including GPUs and TPUs, and is well-suited to machine learning, data science, and education. We can use Colab to manage collections using Supabase Vecs.
In this tutorial we'll connect to a database running on the Supabase platform. If you don't already have a database, you can create one here: database.new.
Start by visiting colab.research.google.com. There you can create a new notebook.
We'll use the Supabase Vector client, Vecs, to manage our collections.
At the top of the notebook, paste the following code and click "Execute" (ctrl+enter):
pip install vecs
On your project dashboard, click Connect. The connection string should look like postgres://postgres.xxxx:[email protected]:6543/postgres
Create a new code block below the install block (ctrl+m b) and add the following code using the Postgres URI you copied above:
import vecs
DB_CONNECTION = "postgres://postgres.xxxx:[email protected]:6543/postgres"
# create vector store client
vx = vecs.create_client(DB_CONNECTION)
Execute the code block (ctrl+enter). If no errors were returned then your connection was successful.
Now we're going to create a new collection and insert some documents.
Create a new code block below the install block (ctrl+m b). Add the following code to the code block and execute it (ctrl+enter):
collection = vx.get_or_create_collection(name="colab_collection", dimension=3)
collection.upsert(
vectors=[
(
"vec0", # the vector's identifier
[0.1, 0.2, 0.3], # the vector. list or np.array
{"year": 1973} # associated metadata
),
(
"vec1",
[0.7, 0.8, 0.9],
{"year": 2012}
)
]
)
This will create a table inside your database within the vecs schema, called colab_collection. You can view the inserted items in the Table Editor, by selecting the vecs schema from the schema dropdown.
Now we can search for documents based on their similarity. Create a new code block and execute the following code:
collection.query(
query_vector=[0.4,0.5,0.6], # required
limit=5, # number of records to return
filters={}, # metadata filters
measure="cosine_distance", # distance measure to use
include_value=False, # should distance measure values be returned?
include_metadata=False, # should record metadata be returned?
)
You will see that this returns two documents in an array ['vec1', 'vec0']:
It also returns a warning:
Query does not have a covering index for cosine_distance.
You can lean more about creating indexes in the Vecs documentation.