Hex

Role: You are an SQL engineer writing a warehouse-efficient query for Hex. Constraints: 1) Input table name events with columns user_id, event_time (timestamp), event_name; 2) Use a single SQL cell, avoid CTEs that prevent pushdown when possible; 3) Week starts on Monday, UTC. Output format: Provide a single parameterized SQL query that returns week_start (date), weekly_active_users, daily_active_users (optional column showing max DAU in that week). Example comment: -- replace events with my_schema.events. Include brief one-line explanation of counting approach and performance tip.

Role: You are a product analyst preparing a reusable Hex SQL cell. Constraints: 1) Provide a safe, parameterized SQL template using {{start_date}} and {{end_date}} placeholders; 2) Validate casting and default values so empty parameters fall back to last 30 days; 3) Include LIMIT and ordering to keep quick previews. Output format: Return only the ready-to-paste SQL with a short comment header explaining parameters and defaults. Example: -- start_date: 2024-01-01, end_date: 2024-01-31. No extra prose, only SQL and header comments.

Role: You are a product analyst building a 12-week retention cohort for Hex. Constraints: 1) Input table events(user_id, event_time, event_name); cohort defined by first purchase event named 'purchase'; 2) Output weekly cohorts for 12 weeks post-cohort; 3) Optimize for pushdown aggregation. Output format: Provide (A) SQL that produces columns: cohort_week_start, week_number, users_in_cohort, returning_users, retention_rate; (B) Example Hex Python/SQL cell snippet that renders a heatmap with cohort_week_start on y and week_number on x; (C) brief note about index/partition recommendations. Include one small example row set.

Role: You are an analytics engineer authoring a repeatable ETL test suite for Hex-run validations. Constraints: 1) Accept a single variable table_name; 2) Produce 6 tests: row_count change threshold, null-rate per critical column, unique key violation, referential integrity sample, column type drift, and max-lag check; 3) Format tests so they can be scheduled nightly and return rows when failing. Output format: JSON array where each element has keys: test_name, sql_query, failure_condition_description, severity. Include one concrete example using table_name=my_schema.orders and thresholds.

Role: You are an analytics/security engineer drafting row-level permission policies for Hex dataset access. Instructions: 1) Given roles: admin, data_analyst, sales_rep, and finance, and sample table schema orders(order_id, account_id, region, amount, owner_id), provide SQL filter expressions per role; 2) Admin sees all rows; data_analyst sees region IN (...) OR owner_id IS NULL; sales_rep sees owner_id = CURRENT_USER_ID placeholder; finance sees amount > 0 and region IN finance_regions; 3) Output format: JSON with role, sql_filter_expression, example_evaluation (one sample row and whether it would be visible). Provide two few-shot examples demonstrating evaluation. Also supply deployment steps for Hex dataset row-filter configuration.

Role: You are a data scientist building a Hex app to explore model predictions with live parameters. Multi-step instructions: 1) Provide a multi-cell Hex notebook blueprint: SQL cell to fetch feature rows with {{sample_date}} and {{customer_id}} parameters; Python cell to load model artifact from S3 and run predictions; Python cell to compute SHAP-like feature importances for a selected row; visualization cell to show prediction, probability, and feature contribution bar chart; deployment cell to expose parameters as app inputs. Output format: return numbered cells with code, parameter declarations, and a small example using customer_id=123 and sample_date=2025-01-01.

Choose Hex over Mode if you need notebook-style SQL+Python collaboration and embeddable parameterized apps for internal tooling.