Developer Center

Resources to get you started with Algorithmia



Algorithms can easily access databases hosted on the Snowflake data platform using the Snowflake Connector for Python. To see this in action on Algorithmia, check out the SnowflakeAsyncOrchestrator algorithm, which provides a reference architecture for performing an asynchronous database write operation. If you’d like to get started with a less complex example, you can follow along below. Note that to complete the following, you’ll need an active account on Snowflake.

To begin, create a hosted data collection named SnowflakeCredentials under the ownership of your own Algorithmia account and upload a file credentials.json with the following structure (see the Snowflake docs to find your Account Name). This file contains session-level parameters that’ll enable you to establish a connection with your Snowflake instance.

  "password": "SNOWFLAKE_PASSWORD",
  "account": "SNOWFLAKE_ACCOUNT",
  "application": "Algorithmia"

Next, create a generic Python algorithm under the ownership of your account. Click “Dependencies” in the Web IDE (or edit your requirements.txt file) and add the dependency snowflake-connector-python.

Now paste in the following as your algorithm code, replacing ALGORITHMIA_ACCOUNT_NAME with your account name so that it will point to the correct data collection and have the appropriate access permissions.

import Algorithmia
import snowflake.connector

# Get Snowflake credentials.
client = Algorithmia.client()
creds = client.file("data://ALGORITHMIA_ACCOUNT_NAME/SnowflakeCredentials/credentials.json").getJson()

def apply(input):
    con = snowflake.connector.connect(
    cs = con.cursor()
        cs.execute("USE SNOWFLAKE_SAMPLE_DATA")
        cs.execute("SELECT * FROM TPCDS_SF100TCL.CALL_CENTER")
        one_row = cs.fetchone()
        return [str(i) for i in one_row]

Build and test the algorithm. Assuming you haven’t deleted the SNOWFLAKE_SAMPLE_DATA database from your Snowflake account, it should return the first row from the sample CALL_CENTER table. Once you verify that the connection is working, you can modify the code to point at your own database tables as appropriate.