AV-SQL: Decomposing Complex Text-to-SQL Queries with Agentic Views

AV-SQL represents a classic research-led 'prompting pattern' for Text-to-SQL. While the concept of 'Agentic Views' is a clever way to handle large schemas (by creating intermediate abstractions), the project lacks a structural moat. With 0 stars and 6 forks at 9 days old, it is currently a niche research artifact rather than a tool with developer momentum. The defensibility is low because the technique—decomposing queries and using LLMs to define temporary views—is a strategy that can be (and is being) replicated by enterprise data platforms like Snowflake (Cortex), Databricks (AI Functions), and specialized startups like Vanna.ai or Dataherald. Frontier labs (OpenAI/Google) are also baking improved SQL reasoning directly into their model capabilities and system prompts. The 'Agentic View' approach is more likely to be absorbed as a standard best practice within larger SQL-agent frameworks than to survive as a standalone product or high-value library.