What exactly is conversational BI?

Conversational BI is a category of analytics technology where users interact with their data by asking questions in natural language rather than writing SQL or navigating dashboards. The system interprets the question, generates the appropriate database query, executes it, and returns results â€” often with a visualization. The term was defined by Gartner as part of the broader augmented analytics trend, representing the next evolution beyond traditional dashboard-based business intelligence.

How is Nova AI different from using ChatGPT to write SQL?

The critical difference is schema awareness . ChatGPT and similar general-purpose AI models do not have access to your actual database schema. They generate plausible-looking SQL based on training data, but the table names, column names, and relationships they reference may not exist in your database. Nova connects directly to your real schema through the Query Streams Agent, so every generated query references actual tables and columns. Nova also generates dialect-correct SQL specific to your database engine, validates queries before execution, and maintains conversation context for multi-step analysis.

Which databases does Nova support for conversational BI?

Nova generates dialect-correct SQL for PostgreSQL, SQL Server, MySQL, MariaDB, Oracle, Snowflake, Google BigQuery, SQLite, DuckDB, and Microsoft Access . Each database engine has distinct SQL syntax rules, and Nova automatically adapts its output to match the target dialect. All of these databases can be queried through the same Nova conversational interface â€” users never need to think about which engine their data lives on.

Do I need to manually describe my database schema to Nova?

No. Nova's Schema Intelligence feature reads your database schema directly through the Query Streams Agent. When you connect a data source, Nova automatically discovers tables, columns, data types, primary keys, and foreign key relationships from the database engine's system catalogs. You do not need to upload schema files, paste DDL statements, or manually describe your data model. The schema is read live, so it always reflects the current state of your database.

What types of charts can Nova generate automatically?

Nova supports six chart types : line charts, bar charts, pie charts, scatter plots, area charts, and horizontal bar charts. Nova selects the most appropriate chart type based on the data structure returned by the query â€” time-series data defaults to line charts, categorical comparisons default to bar charts, and so on. You can switch between chart types after execution if you prefer a different visualization of the same results.

How does the Nova credit system work?

Nova operates on a credit-based system where each AI-powered query consumes credits from your account balance. This gives organizations predictable cost control over conversational BI usage rather than unpredictable per-seat licensing. Credits can be purchased through the Query Streams portal , and usage is tracked per user for full visibility into consumption patterns. Free-tier accounts include credits to get started.

Can Nova handle complex queries with JOINs and subqueries?

Yes. Because Nova reads your full schema including foreign key relationships, it can generate multi-table JOIN queries, subqueries, aggregations, GROUP BY clauses, window functions , and other complex SQL patterns. Nova's accuracy improves with well-defined schemas â€” databases with clear foreign key relationships and descriptive column names produce the best results. For extremely complex analytical queries requiring CTEs or stored procedures, you can always use Query Streams' saved query system for manual SQL.

Is my database data sent to external AI servers?

Nova sends your database schema metadata (table names, column names, data types) to the AI model to generate SQL. The actual data rows from your database are not sent to the AI. The generated SQL is executed locally by the Query Streams Agent against your database, and results are streamed directly to your browser through the secure Query Streams relay. Your underlying data stays within the Query Streams infrastructure and is never exposed to third-party AI training pipelines.

Can I save and reuse queries that Nova generates?

Nova maintains a full conversation history , so you can revisit previous questions and their results at any time. For queries that become recurring reports, you can use Query Streams' saved query system to create a permanent, shareable query based on the SQL Nova generated. That saved query then becomes available in the Excel add-in and Google Sheets add-on for any authorized user to run on demand â€” with interactive filters and parallel execution support.