Snowflake account managers can use the Snowflake connector to directly connect their Snowflake account to Akkio, making it easier than ever to build and deploy AI models on Snowflake data.
To connect your Snowflake account, you simply need to select Snowflake in the primary input step when creating a new flow.
If you don’t already have a Snowflake account, you can sign up for a Snowflake Free Trial. After signing up for Snowflake, you get a custom Snowflake URL. Simply paste in the snippet that comes before “snowflakecomputing.com,” as seen below.
Then, in Snowflake, you’ll want to ensure that your account role is set to ACCOUNTADMIN.
Next, you’ll want to enter any databases you want to give Akkio read access to in your Snowflake account. Multiple database names should be separated by commas.
Now, Akkio will give you a code snippet to paste into Snowflake.
In Snowflake, you’ll first paste in the code, then check the box that says “All Queries,” and then hit “Run.” At the bottom of the page, you’ll get a JSON Blob with OAUTH_CLIENT_ID and OAUTH_CLIENT_SECRET.
Note: We default to using the data warehouse
COMPUTE_WH. The default snippet will provision that warehouse to the new Akkio user.
The final step looks like the below. After pasting in the aforementioned JSON blob (with OAUTH_CLIENT_ID and OAUTH_CLIENT_SECRET), click “Finish.”
You’ll then be automatically taken to a Snowflake page to verify your integration. Simply click “Allow,” and you’re done!
From here, you can create an Akkio flow in the same way you would with any other data source. Simply select your Snowflake dataset, and you can either merge with additional data, or go straight to modeling and deployment.
For this example, we’ll use a restaurant review dataset that’s located on Snowflake, which has just two columns: “Reviews,” which are strings of restaurant reviews, and “Liked,” the binary column of “0” or “1” that we’re trying to predict. Since we’ve integrated Snowflake, we can simply search for and select this dataset.
Again, we can make predictions in the same way now as any other flow. Let’s hit “Add Step” and then the “Liked” column to predict whether a review has positive or negative sentiment.
In seconds, we’ll have a highly accurate model to predict restaurant review sentiment.
Finally, we can deploy our model in virtually any environment. Since we’ve already integrated with Snowflake, let’s deploy our model back to Snowflake. What this means is that we can automatically make predictions on new data in Snowflake. Simply hit “Add Step,” and then select Snowflake in the “Outputs” section.
The final step is to simply fill out the deployment settings, and ensure that the fields are mapped correctly to the new data. That’s it! When you’re done, you can hit “Show Preview” and “Deploy.”