Snowflake Cortex AI
Product · API / developer platform · Engineering & development by Snowflake
Securely build and deploy enterprise AI agents grounded in your data.
A fully managed suite of AI services for building generative AI apps and agents. It provides secure access to leading LLMs and high-accuracy data analysis tools directly within the Snowflake data perimeter.
Category: Generative AI
- APIs and SDKs
- SQL, Python API, Cortex REST API, Snowflake Python APIs
- Programming Languages
- SQL, Python
- Deployment Model
- Fully Managed Serverless
- Primary Interfaces
- Snowsight, REST API, Python SDK
- Data Formats
- Structured, Semi-structured, Unstructured (Text, PDF, Video)
- Governance Framework
- Snowflake Horizon Catalog Integration
- Access to Llama 3.1, Mistral Large, and Snowflake Arctic LLMs
- High-accuracy text-to-SQL for conversational data analysis
- Fully managed serverless infrastructure with no GPU management required
- Unified governance and observability for all AI agents and models
Overview
Snowflake Cortex AI is a suite of fully managed AI services designed for businesses to build and deploy generative AI applications securely. The platform enables high-accuracy text-to-SQL queries through Cortex Analyst and enterprise-grade search for unstructured data via Cortex Search. By integrating AI directly into the Snowflake perimeter, it ensures that proprietary data remains governed by the same security and access controls as the rest of the Data Cloud. Cortex AI also includes tools for model fine-tuning, observability, and centralized agent governance through the Cortex AI Gateway.
Generative AI Services
- Cortex Analyst: Build high-accuracy text-to-SQL services that allow business users to chat with structured data using natural language.
- Cortex Search: Create enterprise search and retrieval-augmented generation applications on unstructured text data with scalable retrieval.
- Cortex Training: Securely fine-tune and post-train open models using proprietary enterprise data to create specialized intelligence.
Governance and Controls
- Cortex AI Gateway: Govern every AI agent across the business with centralized controls for model routing, usage tracking, and budget management.
- Unified Governance: Apply unified security policies and access controls to both data and AI workloads within a single compliance perimeter.
- Cost Management: Monitor AI credit consumption by user, team, and workload with native tagging for precise cost attribution.
Application Development
- Multimodal Processing: Analyze text, documents, and other media at scale using multimodal AI capabilities integrated into the data platform.
- Automatic Data Agents: Automatically generate AI-powered agents and semantic views for data listings to enable conversational querying.
- Developer Experience: Develop AI applications using SQL or Python through a serverless architecture that requires no GPU infrastructure management.
- Streamlit Integration: Build user interfaces for AI apps using Streamlit, an open-source Python library for sharing custom applications.
Use cases
- Business users query structured databases using natural language to generate accurate SQL and insights without technical expertise.
- Developers build RAG-enabled conversational assistants that search through years of animal care records or technical documents.
- Data engineers automate the extraction of sentiment and key terms from large volumes of unstructured customer feedback or SEC filings.
- Organizations fine-tune open-source models on proprietary datasets to create specialized AI for industry-specific tasks like claims management.
Specifications
- APIs and SDKs
- SQL, Python API, Cortex REST API, Snowflake Python APIs
- Deployment
- Fully managed serverless cloud platform (AWS, Azure, Google Cloud)
- Supported LLMs
- Llama 3.1/3.2, Mistral Large, Snowflake Arctic, Reka Core, Gemma-7b
- Programming Languages
- SQL, Python
- Deployment Model
- Fully Managed Serverless
- Primary Interfaces
- Snowsight, REST API, Python SDK
- Data Formats
- Structured, Semi-structured, Unstructured (Text, PDF, Video)
- Governance Framework
- Snowflake Horizon Catalog Integration
- Cost Control
- Per-user quotas and AI credit consumption tracking
- Search Capabilities
- Vector search and Retrieval-Augmented Generation (RAG)
How it works
- Prepare Data Foundation: Connect your structured and unstructured data to the Snowflake AI Data Cloud to establish a governed foundation.
- Select or Train Models: Select from industry-leading LLMs or fine-tune open models using Cortex Training within the secure Snowflake perimeter.
- Build with SQL or Python: Use SQL or Python to call Cortex AI functions for tasks like text generation, translation, or document analysis.
- Deploy and Govern: Apply governance policies through Cortex AI Gateway and monitor performance with built-in observability tools.
FAQ
What is the Cortex AI Gateway?
Cortex AI Gateway brings every agent under a unified set of controls, allowing enterprises to secure AI connections, track usage, and manage AI spending through dynamic model routing.
What is the difference between Cortex Analyst and Cortex Search?
Cortex Analyst is a high-accuracy text-to-SQL service that allows users to chat with structured data, while Cortex Search builds RAG applications on unstructured text data.
Can I fine-tune models with my own data?
Yes, Cortex Training allows organizations to securely fine-tune and post-train open models using their own enterprise data within the Snowflake security perimeter.
Added by AntSpace from public sources (snowflake.com).