I was quietly working on a tool that connects to BigQuery and many more integrations and runs agentic analysis to answer complex "why things happened" questions.
It's not text to sql.
More like a text to python notebook. This gives flexibility to code predictive models on top of bigquery data as well as react data apps from scratch.
Under the hood it uses a simple bigquery lib that exposes query tools to the agent.
The biggest struggle was to support environments with hundreds of tables and make long sessions not explode from context.
It's now stable, tested on envs with 1500+ tables.
Hope you could give it a try and provide feedback.
matankleym•1h ago
Under the hood it uses a simple bigquery lib that exposes query tools to the agent. The biggest struggle was to support environments with hundreds of tables and make long sessions not explode from context.
It's now stable, tested on envs with 1500+ tables. Hope you could give it a try and provide feedback.
TLDR - Agentic analyst connected to BigQuery