<aside> 💡

You may review the full installation script with comments here. If you have any question, get in touch with your CSM or write us at [email protected]

</aside>

Overview

This document provides an overview of Lang.ai's Native App technical design, focusing on data security and privacy.

Data Residency Lang.ai's Native App is designed with a "data-stays-put" principle. By default, all your data remains securely within your Snowflake instance, ensuring maximum control and compliance with data governance policies.

Security Features

  1. Built on Snowflake's Security Infrastructure Lang.ai's Native App utilizes Snowflake's networking configuration and relies on Snowflake's Role-Based Access Control (RBAC) for authentication and authorization. Learn more here.
  2. Minimal Outbound Connectivity No outbound rules are required for core functionality. Outbound connectivity is only needed for the optional Slack integration.
  3. In-House Language Model The app leverages Llama 3.1 405 from Snowflake's Cortex suite, eliminating dependency on external AI services.
  4. Controlled Logging Lang.ai uses Snowflake's native logging functionality. Logging is disabled by default. When enabled, all logs are inspectable by the customers. For more information on enabling logs, see here.

These measures ensure that your team maintains full control over data movement outside of Snowflake, preventing unauthorized data transmission even as our code runs in its own container.

Architecture

lang-architecture.png

Use Case Flow

  1. Grant Relevant Data access to the AI Agent
  2. Create View to join all Relevant Data
  3. AI Agent runs analysis automatically based on the data in the View
  4. Insights are generated for a specific period of time or automatically on a recurring basis