Use the query assistant

The assistant turns a description of the records you want into a saved relationship query. Manual queries work without an AI key.

Configure a provider

For local development, copy the repository’s .env.example to .env.local and set AI_PROVIDER, AI_MODEL, and the matching key. Use a model that supports tool calling. AI_MODEL is the provider’s model ID without a provider prefix.

AI_PROVIDER	Key variable
openai	OPENAI_API_KEY (or OPEN_AI_API_KEY)
anthropic	ANTHROPIC_API_KEY
google	GOOGLE_GENERATIVE_AI_API_KEY

Packaged apps read these settings from the process environment. Development loads the root .env.local without overriding existing environment variables. Restart Electron after changing the configuration.

Describe the question

Select a project and connection, then describe the base records, related information, filters, and sorting you want. For example:

Find the customer with email alex@example.com and show their orders, newest first. Include the customer email, order ID, and order date.

Answer any clarification in the chat. Once the query is saved, use the same conversation to amend it. Unsaved manual edits are included in the next chat amendment.

The tool validates its proposed query against the schema and compiles it with the query engine before saving. Review the resulting tables and filters to check they answer your question.

Understand what is sent

The provider receives the database schema, conversation, and current query, including filter values. It does not receive result rows or database connection credentials. API keys remain in Electron’s main process.

Only enable a provider appropriate for the data in your schema and query filters. Assistant updates save the query and conversation automatically; clarification before an update stays in memory.

The assistant supports the relationship builder’s read queries. Use the SQL workspace for custom SQL.