Guides AI agents through Wren CLI workflows for querying databases via the MDL semantic layer.
npx skills add https://github.com/Canner/WrenAI --skill usage-canner
Please help me install this Agent Skill. Skill: usage Source: https://github.com/Canner/WrenAI/tree/main/core/wren/src/wren/skills_content/usage Command: npx skills add https://github.com/Canner/WrenAI --skill usage-canner
wren memory fetch, recall past queries as few-shot examples, write dialect-neutral SQL against MDL models, execute, and store confirmed results. - Structured error recovery: Diagnose failures with wren dry-plan to separate MDL-level issues (wrong model or column names) from database-level issues (type mismatches, permissions, timeouts). - Cube query support: Answer aggregation questions like "revenue by month" or "top N customers" through wren cube query instead of hand-writing GROUP BY SQL. - Use Case: A user asks "what were our top 5 customers by revenue last quarter?" The agent recalls similar past queries, fetches schema context, checks for a matching cube, executes the query through the semantic layer, and stores the confirmed result for future sessions. ## Quick Start Ask the agent to answer a business data question such as "show me monthly revenue for 2024" using the wren CLI and let it follow the guided workflow.High-intent search queries and answers about installing and using this skill.
Fetch schema context with wren memory fetch, recall similar past queries with wren memory recall, then write SQL against MDL model names and execute with wren --sql. Store successful queries with wren memory store so future sessions can reuse them.
Run wren dry-plan with the failed SQL to separate MDL-level errors from database-level errors. If dry-plan fails, check model and column names with wren memory fetch; if it succeeds but execution fails, compare the expanded SQL against the database error message.
Wren supports postgres, mysql, bigquery, snowflake, clickhouse, trino, mssql, databricks, redshift, spark, athena, and oracle through pip extras, with DuckDB included by default. Install the matching extra such as pip install "wrenai[postgres]".
Use wren cube query for aggregation questions like totals by month or top-N rankings when the MDL defines cubes covering the question. Fall back to wren --sql for custom JOINs, window functions, CTEs, subqueries, or when no cubes are defined.
Do not store queries that failed, returned results the user rejected, or are exploratory such as SELECT * with LIMIT. The CLI auto-detects exploratory queries, and raw SQL without a natural language question should also be skipped.
CannerAutomate Wren Engine workflows for SQL, MDL, and MCP server tasks.
Datus-aiValidate BI dashboard configurations and data accuracy for Superset and Grafana.
AradotsoBuild a semantic context layer for AI agents querying data warehouses.
kortix-aiWrite dialect-specific SQL for PostgreSQL, Snowflake, BigQuery, and Redshift.