Skip to main content

​
Overview

The Snowflake Cortex REST API integration allows Genesis Data Agents to leverage state-of-the-art large language models (LLMs) directly within your Snowflake environment. This enables powerful AI capabilities for text generation, embeddings, and unstructured data analysis—all without requiring a Snowflake warehouse.

​
Key Benefits

  • No Warehouse Required: Cortex REST API operates without warehouse compute costs
  • Multiple Model Options: Access models from OpenAI, Anthropic, Meta, Mistral AI, and DeepSeek
  • Server-Sent Events: Streaming token generation for responsive user experiences
  • Native Integration: Seamlessly works within Genesis agent implementations

​
How Genesis Uses Cortex REST API

Genesis integrates with Cortex REST API through the bot_implementation configuration, allowing agents to use Cortex models as their underlying LLM provider.
The Team Wiki Librarian uses your active LLM provider. When Cortex is the active provider, set wiki.extractor_model to a Cortex model such as cortex/claude-sonnet-4-6 (pick it from the dropdown in Advanced Settings). If you later switch the active provider away from Cortex, the Librarian stops calling Cortex until you choose a model from the new provider. See Team Wiki Learnings.

​
Prerequisites

Before configuring Cortex REST API in Genesis, ensure you have:
  1. Snowflake Account with Cortex AI features enabled in a supported region
  2. Authentication Credentials: Programmatic Access Token (PAT) or Snowflake key-pair (PEM private key plus account and user)
  3. CORTEX_USER Role: Granted to your user/role (usually inherited from PUBLIC role)
  4. Genesis Data Agents platform deployed (SPCS or standalone)

​
Configure in Genesis

On standalone or non-native deployments, open Config -> AI Providers, select Snowflake Cortex, and choose Authentication Method:
  • Personal Access Token (PAT) — enter Account URL and Personal Access Token.
  • Key Pair (JWT) — enter Snowflake Account, Snowflake User, Private Key (PEM), and optional Private Key Passphrase for encrypted keys.
Use Test Connection to verify either mode before saving. In Snowflake Native App / SPCS runtimes where Cortex is already available, Genesis can use native runtime credentials without entering these fields. See AI Providers for embedding-method options and model-assignment details.

​
Additional Resources

​
Support

For issues specific to:
  • Cortex REST API: Contact Snowflake Support
  • Genesis Integration: Email [email protected]
  • Rate Limit Increases: Submit ticket to Snowflake Support with usage projections