Snowflake Integration

Datafi connects to Snowflake through a native driver that supports both JWT and OAuth authentication. Once connected, you can query Snowflake data alongside other sources using Datafi's federated query engine. Queries authored in PRQL are compiled to Snowflake-compatible SQL before execution.

Authentication Methods

Datafi supports two authentication methods for Snowflake. Choose the one that aligns with your organization's security requirements.

JWT (Key Pair Authentication)

JWT authentication uses an RSA key pair. Snowflake validates the JWT token signed with your private key against the public key registered to your Snowflake user.

Setup steps:

Generate an RSA key pair (2048-bit minimum):

openssl genrsa 2048 | openssl pkcs8 -topk8 -inform PEM -out rsa_key.p8 -nocrypt
openssl rsa -in rsa_key.p8 -pubout -out rsa_key.pub

ALTER USER datafi_service_user SET RSA_PUBLIC_KEY='MIIBIjAN...';

In Datafi, navigate to Integrations > Add Data Source > Snowflake.
Select JWT as the authentication method.
Provide your Snowflake account identifier, username, and upload the private key file.
Click Test Connection, then Save.

OAuth

OAuth authentication delegates credential management to your identity provider. Datafi acts as an OAuth client and obtains access tokens through the authorization code flow.

Setup steps:

CREATE SECURITY INTEGRATION datafi_oauth
  TYPE = OAUTH
  OAUTH_CLIENT = CUSTOM
  OAUTH_CLIENT_TYPE = 'CONFIDENTIAL'
  OAUTH_REDIRECT_URI = 'https://home.datafi.us/oauth/callback'
  ENABLED = TRUE;

Retrieve the client ID and client secret from the integration.
In Datafi, select OAuth as the authentication method.
Enter the client ID, client secret, and your Snowflake account identifier.
Click Authorize to complete the OAuth flow.

tip

OAuth is recommended for organizations that require centralized credential rotation and token lifecycle management through their identity provider.

Connection Parameters

Parameter	Description	Required
Account Identifier	Your Snowflake account locator (e.g., `xy12345.us-east-1`).	Yes
Username	The Snowflake user account Datafi uses to connect.	Yes (JWT)
Private Key	The PKCS#8 private key file for JWT signing.	Yes (JWT)
Client ID	OAuth client identifier from the security integration.	Yes (OAuth)
Client Secret	OAuth client secret.	Yes (OAuth)
Database	The default database context for queries.	No
Schema	The default schema within the selected database.	No
Role	The Snowflake role assumed for all sessions.	No

Warehouse Selection

Datafi allows you to select the Snowflake virtual warehouse used for query execution. You configure this at the data source level.

Warehouse Setting	Description
Default Warehouse	The warehouse used for all queries unless overridden.
Auto-Suspend	Datafi respects your warehouse auto-suspend settings. No additional configuration is needed.
Warehouse Size	Choose the size that matches your expected workload (X-Small through 6X-Large).

note

If the specified warehouse is suspended when a query arrives, Snowflake automatically resumes it. Resume time depends on the warehouse size and typically takes a few seconds.

Schema Introspection

When you connect a Snowflake source, Datafi performs automatic schema introspection to catalog all accessible databases, schemas, tables, views, and columns. This metadata powers the Data Catalog, autocomplete in the query editor, and policy enforcement.

Introspection runs:

On initial connection -- A full scan of all accessible objects.
On schedule -- Periodic refresh (configurable, default every 6 hours).
On demand -- You can trigger a manual refresh from the data source settings.

PRQL Compilation to Snowflake SQL

Datafi uses PRQL (Pipelined Relational Query Language) as its intermediate query representation. When you write a query in Datafi -- whether through SQL, natural language, or a Data View -- the Coordinator compiles it to PRQL and then to Snowflake-compatible SQL.

Example PRQL to Snowflake SQL:

from orders
filter order_date >= @2024-01-01
derive total = quantity * unit_price
sort {-total}
take 100

Compiles to:

SELECT *, quantity * unit_price AS total
FROM orders
WHERE order_date >= '2024-01-01'
ORDER BY total DESC
LIMIT 100;

The PRQL compiler handles Snowflake-specific syntax -- including QUALIFY, FLATTEN, semi-structured data access, and Snowflake SQL functions -- automatically.

Troubleshooting

Issue	Possible Cause	Resolution
`JWT token is invalid`	Public key mismatch or expired token.	Re-register the public key with the Snowflake user.
`Warehouse is not accessible`	The Datafi role lacks `USAGE` privilege on the warehouse.	Grant `USAGE` on the warehouse to the configured role.
`Schema introspection timeout`	Large number of objects or network latency.	Narrow the default database/schema or increase the introspection timeout.

Next Steps

PostgreSQL and MySQL -- Connect relational databases.
Cloud Warehouses -- Set up BigQuery, Databricks, Redshift, or Synapse.

Authentication Methods​

JWT (Key Pair Authentication)​

OAuth​

Connection Parameters​

Warehouse Selection​

Schema Introspection​

PRQL Compilation to Snowflake SQL​

Troubleshooting​

Next Steps​