AI Smart Assistance

Team Edition offers the ability to construct SQL queries using natural language with AI smart completion and AI Chat. This feature works through integrations with the following AI services:

• OpenAI (GPT language model)
• Azure OpenAI (Azure OpenAI Service)
• Google Gemini (Gemini)
• Ollama (Ollama)

Team Edition is not affiliated with OpenAI, Microsoft Azure, Google, or Ollama. Integration with AI assistance is achieved through the public APIs. To use AI assistance, register with the respective service provider and get a secret key.

To activate the AI assistance in Team Edition, configure the API token:

1. Navigate to Window -> Preferences -> General -> AI.

2. Ensure the Enable smart completion option is activated.

3. In the API token field, input your AI secret key.

4. Save the changes.

   

   For more info on configuration settings, see AI assistant settings.

• AI smart completion. Type plain-language queries, and AI converts them into SQL.

• AI Chat. Chat naturally with AI to get your queries.

• AI command. Use AI command directly in the SQL Editor. Type commands like @ai show films in which Grace Mostel starred, and AI wll generate the SQL for you.

Best practices for prompt formulation:

• Language: While AI supports multiple languages, it's recommended to use English for best results.
• Database knowledge: Familiarity with your database structure enhances the accuracy of generated queries.
• Iterative refinement: Start with a general request and refine iteratively. As you provide more specific details, the AI's output will become more relevant.
• Explicit details: If you know certain tables or columns that should be part of the query, include them in your request for better accuracy.

We prioritize data safety and user privacy. In this section, we outline how data is managed and the measures taken to protect user privacy when using the AI assistance.

To enable the AI assistance capabilities, table and column names from the current database schema are transmitted to OpenAI. This step is crucial for accurately translating user requests into SQL queries. Key considerations regarding data privacy are as follows:

• No Table Data: Only metadata like table and column names are shared with OpenAI. Actual table data is not transmitted.

• User Consent: On the first use of AI completion for a specific connection, Team Edition will prompt for your confirmation to send metadata. This confirmation is mandatory to use the feature.

  

• Log Transparency: The entire request can be logged for your review. To enable this, navigate to Preferences and check the Write GPT queries to debug log option.

• Selective Metadata Sharing: If you prefer not to share information about certain tables, adjust the tables in scope using the Scope field.
• Azure OpenAI privacy: If you use Azure OpenAI, be aware that it operates under its own privacy policy. It's recommended to review their terms before using.

• Google Gemini privacy: When utilizing Google Gemini, it is important to understand the specific data privacy measures.

Learn more about disabling AI assistance.