Creating a Snowflake Query Assistant

An exciting use-case we get asked about is the ability to query snowflake for relevant information. Building an assistant that can query and analyze snowflake data allows many people in your organization to get insights they normally would struggle to get. Let’s build this out!

Step 1: Figure out what questions you want to ask

Think about what metrics and analytics are requested most often, and what questions you’d like to get answered with data.

1. Get Recent Data: Query to get up to date data on current metrics stored in a snowflake table
2. Analyze Data: Let the snowflake agent query data and then run code to extract metrics.
3. Queries: Allow the LLM to generate queries or run pre-defined query templates.

The snowflake action is quite powerful in unlocking new to users who wouldn’t otherwise have it.

Step 2: Planning Your Approach

Before building, consider these key questions:

1. What tables and views does your agent need access to? Think about sensitive data and what you’d like the agent users to query
2. Who will maintain it? Identify someone with Snowflake access who will keep the Agent updated with new views or query templates depending on your approach (will be explained in the next steps)

The beauty of Credal Agents is that you can start with a simple implementation and enhance it over time as you learn what works best for your team.

Ready to build Version 1 of your snowflake assistant? Let’s get started with a initial setup to get you started!