GitHub user mergisi added a comment to the discussion: Vambery AI Agent — AI-powered SQL assistant extension for SQL Lab (public beta)
Really interesting to see text-to-SQL integrated directly into Superset's SQL Lab via the Extension System. The schema-awareness piece is what separates useful NL-to-SQL from toy demos — knowing which tables, columns, and relationships exist before generating a query eliminates the most common failure mode (hallucinated column names). A few questions on the architecture: 1. **Query validation layer** — does Vambery validate the generated SQL against the schema before execution, or does it rely on the database to catch errors? A pre-execution AST check could save round-trips and give better error messages to users. 2. **Multi-dialect handling** — Superset supports many database backends. How does the agent handle dialect-specific syntax (e.g., `LIMIT` vs `TOP`, date functions across Postgres/MySQL/BigQuery)? 3. **Context window management** — for databases with hundreds of tables, how do you decide which schema context to include in the prompt? Full schema injection doesn't scale, so curious if you're doing relevance filtering or embedding-based retrieval. The `set_editor_sql` + auto-execute flow is a nice UX touch — it keeps the user in control while reducing friction. Disclosure: I'm working on [ai2sql.io](https://ai2sql.io), a standalone natural language to SQL tool. Always great to see more projects pushing this space forward, especially ones embedded directly in the analytics workflow like this. GitHub link: https://github.com/apache/superset/discussions/38356#discussioncomment-15982938 ---- This is an automatically sent email for [email protected]. To unsubscribe, please send an email to: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
