Sure! Here’s the translation into American English:
Artificial Intelligence agents are designed to optimize business productivity by automating complex tasks. However, their effectiveness largely depends on access to quality data, both structured and unstructured. Managing access control, adhering to strict privacy protocols, and retrieving information efficiently remain significant challenges for many organizations.
To address these challenges, Snowflake has developed Cortex Agents, a fully managed service that facilitates the integration, retrieval, and processing of data. With this solution, businesses can develop AI agents that are more accurate and scalable, ensuring efficient access to information without compromising data security or privacy.
Cortex Agents: Bringing AI to Businesses
Cortex Agents, now available in public preview, orchestrate structured and unstructured data sources, whether Snowflake tables or PDF files stored in object storage, to provide valuable insights. They break down complex queries, retrieve relevant data, and generate accurate responses using Cortex Search, Cortex Analyst, and LLM, thereby ensuring precision, efficiency, and governance at every step of the process.
The Cortex Agents schedule tasks, utilize tools to execute them, and reflect on the results to improve responses. Available as a convenient REST APIAn API, short for “Application Programming Interface”, Cortex Agents can seamlessly integrate into any application.
Cortex Analyst: AI-Powered SQL Generation with Semantic Understanding
Cortex Analyst can be utilized as a tool within Cortex Agents.
Unlike typical text-to-SQL systems that rely solely on pattern matching, Cortex Analyst employs a semantic model to map business terms to underlying data. This unique approach enhances accuracy in real-world use cases involving complex multi-table environments.
What’s New in Cortex Analyst
- Handling Greater Schema Complexity: Cortex Analyst now goes beyond star-schema and Snowflake-schema JOINs. Our new advanced JOIN validation mitigates common issues, such as JOIN hallucinations and double counting, that often arise in complex queries. This allows Cortex Analyst to support multi-table queries without compromising accuracy.
- Semantic Model Generation and Monitoring: The public preview of the new Analyst management UI in Snowsight simplifies the process of creating and refining semantic models. Administrators can select tables and columns, and use LLM (operating within Snowflake’s security perimeter) to generate an initial semantic model YAML file. The management interface also monitors user interaction and feedback. This enables clients to track usage and make informed improvements to semantic models over time.
- Customization for Business-Specific Logic: With Custom Instructions now available to the general public, users can tailor Cortex Analyst to their unique business needs using natural language in the semantic model file. Common use cases include specifying fiscal year start dates, explaining internal naming conventions, and prioritizing key tables during SQL generation.
With these updates, Cortex Analyst enhances the analysis of structured data and streamlines management setup for agent-based applications.
Cortex Search: High-Quality Context Engine for Unstructured Data
The Cortex Agents use Cortex Search to retrieve unstructured data (e.g., text, audio, image, video). Cortex Search is a native hybrid search combining vector and lexical (keyword) searches, with an additional step of semantic re-ranking, to provide high-quality, low-latency retrieval at scale.
Cortex Search achieves cutting-edge quality, outperforming competitive enterprise search platforms in retrieval accuracy (NDCG@10) with OpenAI embedding models by at least 11% across a diverse set of benchmark tests.
What’s New in Cortex Search
- Increased Scale and Accessibility: Cortex Search now supports indexing hundreds of millions of rows. Additionally, Cortex Search service costs have been reduced by 30% as a result of infrastructure optimizations.
- Enhanced Customization Capabilities: Cortex Search now offers the ability to select the vector embedding model for semantic search. This includes two multilingual models, snowflake-arctic-embed-l-v2.0 and voyage-multilingual-2. Furthermore, Cortex Search supports filtering by date range in metadata columns.
- New Features in Preview: The new features in preview include the Cortex Search management UI (for observability and quality tuning); boosting and degrading numerical and temporal signals; confidence scoring of results; and advanced filtering capabilities.
With these new features, Cortex Search provides a scalable and customizable foundation for search and agent-based applications using Snowflake data.
AI Observability: Evaluating and Monitoring AI Agents
AI observability brings reliability, performance, and trust to generative AI applications. With proper evaluation and monitoring, businesses can achieve more accurate outcomes, optimize costs, and address governance needs.
What’s New in Cortex AI Observability
Cortex AI observability on Snowflake works with TruLens and will soon be available in public preview.
- Comprehensive Evaluation: AI observability can assess the performance of agents and applications using techniques like LLM-as-a-judge. It can provide metrics such as relevance, grounding, and toxicity, giving customers the ability to quickly iterate and refine the agent to improve performance.
- Comparison: Users can compare evaluation runs side by side and assess the quality and accuracy of responses across different LLM configurations to identify the best configuration for production deployments.
- Full Tracking: Customers can enable logging for each step of the agent’s executions in input prompts, tool usage, and final response generation. This facilitates easy debugging and refinement for accuracy, latency, and cost.
Effective governance and processing of both structured and unstructured data within Snowflake are critical for creating AI-ready datasets that retrieval services can utilize. Snowflake’s support for unstructured data includes capabilities for storing, accessing, processing, managing, governing, and sharing such data. The Snowflake connector for SharePoint ensures that existing permissions are honored to secure access controls. Additionally, Snowflake’s acquisition of Datavolo enhances the platform’s ability to handle multimodal data integration, reinforcing its commitment to strong governance and data processing.
With these capabilities, Cortex AI observability makes AI applications more efficient and reliable for business use.
The Future of AI Agents
AI agents are transcending basic automation, dynamically managing actions and complex reasoning. This represents a significant improvement over current software tools, which are mostly reactive. As LLMs continue to advance, agents will collaborate, plan, execute, and refine tasks, driving efficiency and reducing costs. Agents have the potential to significantly cut expenses, both in software and labor.
Cortex Agents, through Cortex Analyst, Cortex Search, and AI observability, provide intelligence on a unified governance framework and an efficient processing engine for structured and unstructured data.