Snowflake is best described as anAI Data Cloud Platform. While Snowflake originated as a cloud data warehouse, its capabilities have expanded to include secure data sharing, application development, machine learning workloads, unstructured data processing, Snowpark, and integrated LLM/AI functions (such as Snowflake Cortex and Streamlit). This broad ecosystem supports analytic workloads, governance, multi-cloud architecture, and AI-driven applications. It is not simply an ETL tool or a transactional database.

=======================================

CLASSIFY_TEXTis a Snowflake Cortex task-specific function designed to categorize free-form text into predefined buckets. It operates by applying LLM-driven classification logic based on category labels or descriptions provided by the user. A primary and practical use case isautomatically categorizing customer support ticketsbased on content, such as identifying billing issues, login problems, or product defects. This enables automated routing, prioritization, and operational analytics.

The function outputs structured results with the predicted label and confidence score. It is optimized for domain classification rather than open-ended text generation.

Incorrect options:

Generating SQL codeis handled by the COMPLETE function.

Translationis handled by TRANSLATE.

Extracting structured fields from documentsis the role of PARSE_DOCUMENT.

Thus, classification of incoming text streams is the natural use case of CLASSIFY_TEXT.

====================================================

The COPY INTO < table > command is Snowflake’s primary and optimized mechanism for loading large volumes of structured data from either internal or external stages. Unlike INSERT INTO, which is used for row-level inserts or small datasets, COPY INTO is designed for high-throughput operations, parallel loading, and performant ingestion from files stored in S3, Azure Blob Storage, GCS, or Snowflake-managed stages. The COPY INTO command supports detailed file formatting options (such as TYPE=CSV, FIELD_DELIMITER, SKIP_HEADER), error-handling rules (ON_ERROR), and transformations using SQL expressions inside the COPY statement. Snowflake does not support LOAD DATA or IMPORT DATA as SQL commands; these exist in other database systems but not within Snowflake. COPY INTO ensures efficient, resilient, fault-tolerant data ingestion and is heavily used for ETL/ELT, data lake ingestion, and large batch ingestion pipelines.

=======================================

The official Snowflake web interface is calledSnowsight. Snowsight is the modern, browser-based UI that enables SQL development, Python worksheets, dashboarding, visual exploration, and account monitoring. It replaces the classic UI and is now the default interface. Snowsight provides integrated capabilities such as notebooks, data loading interfaces, dashboards, role and warehouse management, and visual query history. Other options like Snowsite, SnowGUI, and SnowView are not valid Snowflake product names.

=======================================

Snowsight provides adrag-and-dropinterface for uploading data from a local machine into an internal stage. This is done via: Data → Add Data → Load Files into a Stage. Users can upload multiple files and choose the destination stage. This convenient UI eliminates the need for command-line tools such as SnowSQL. Snowsight supports both local file uploads and cloud storage references, not just cloud sources.

=======================================

Semi-structured data like JSON stored in a VARIANT column can be queried directly usingpath notation, either dot notation or bracket notation.

Examples:

Dot notation:SELECT data:customer.name FROM table;

Bracket notation:SELECT data[ ' customer ' ][ ' name ' ] FROM table;

Snowflake automatically interprets the JSON structure, making relational extraction unnecessary. Complex fields can be accessed through functions such as FLATTEN, OBJECT_KEYS, ARRAY_SIZE, and TYPEOF.

Incorrect options:

Converting JSON to a relational table is optional, not required.

Stored procedures are not necessary for JSON querying.

WHERE on the full VARIANT cannot precisely extract fields.

Thus, path notation is the native, efficient method.

====================================================

Snowflake’s object hierarchy maintains a structured namespace. At the top is thedatabase, which serves as a container for related schemas. Eachschemaorganizes logical groupings of tables, views, stages, file formats, and other database objects.

Tablesstore persistent, structured data.

Viewsprovide virtual representations of data through stored SQL queries. They do not store data themselves but reference underlying tables.

The hierarchical structure is:

< database > . < schema > . < object >

Incorrect statements:

Views do not contain databases—they reference objects but do not encapsulate them.

Schemas do not manage network or cloud settings; those are defined at account or cloud service layers.

This hierarchy enables clear object governance, secure sharing, and proper namespace organization.

====================================================

External stages point to files in external cloud storage (S3, Azure Blob, GCS). Because the data is not stored inside Snowflake, the user avoids Snowflake storage charges, which can significantly reduce cost for large staging datasets.

External stages do not automatically delete files, are not inherently more secure than internal stages, and still count as database objects even though they reference external storage.

==================

On the Snowsight Schema Details page, one of the supported operations is renaming the schema. The UI exposes controls that allow users with appropriate privileges to change the schema name, which can help maintain consistent naming conventions or reflect project reorganizations.

Sharing data with other accounts is typically done using secure shares at the database or object level, not from a simple “Share this schema” function on the Schema Details page. Truncation is a table-level operation (e.g., TRUNCATE TABLE), not something that applies at the schema level. Metrics visible on the Schema Details page (such as object counts or storage usage) are informational and not directly editable; they are derived from system metadata.

==================

The auto-suspend feature saves compute costs byautomatically suspending a warehouse when it becomes idlefor a configured period. Since Snowflake charges for compute time while a warehouse is running, auto-suspend prevents unnecessary credit consumption by stopping the warehouse when no queries are executing. The warehouse can resume automatically when a new query is submitted, provided auto-resume is enabled. Auto-suspend does not change warehouse size (scaling up/down) nor adjust clusters in a multi-cluster warehouse (scaling in/out). It strictly controls when compute resources turn off due to inactivity, making it an essential cost-optimization feature.

=======================================