Metabase - ChatGPT

Add helper features to Metabase through requests to the OpenAI API

Metabase - ChatGPT brings new features to Metabase through the use of OpenAI's ChatGPT. This extension provides 4 added features to the native query editor of Metabase: - Database error explanation: asks ChatGPT for the most likely explanation for the error considering the query that was run and the error returned - Query prompt: allows you to write a prompt for ChatGPT to create/modify a query as a comment in the query editor - Query reformatting: asks ChatGPT to reformat the query to follow best practices - Error line highlight: if the database error includes the character position of the error, highlights the corresponding line in the query editor (does not rely on ChatGPT) To use most of those features, you need to provide an OpenAI API key linked to a credited account. If you’re interested but do not have an account, I’d recommend you consider getting one as the pay-as-you-go pricing makes it extremely cheap ($0.0005 per request made by this extension in average). Important security information: your API key will be stored in the Chrome storage. So, make sure you're fine with it before using this extension. The the settings of the extension, you have the choice of turning on/off the use of database structure embedding to provide context on your database to Chatgpt. We recommend you turn it on for best results. On your first visit to the query editor page, the extension will create the vector database for the embedding; that can take a while if your database contains a lot of tables. While the circular arrow gif icon is turning, DO NOT close your tab as the embedding is being created. This is a one-off operation, so you won't be bothered by it in the future. This database structure embedding works best if you have defined descriptions of your tables and columns and if you have set foreign keys. This extension is open-source (https://github.com/RobinPicard/metabase-chatgpt). Feel free to help me improve it by creating issues or opening pull requests on Github!