Making the most of working with LLMs in the Valentina Studio SQL Editor
Other Language Versions of this Article: [EN] [ES] [DE] [JP] [RU]
Instructions are system level guidance that are sent to an AI to clarify what you want, and how you want it. All major models such as OpenAI, Gemini, Grok, Anthropic Claude support some form of instruction. They generally support some form of:
- Setting behavior rules
- Defining persona or tone
- Providing general context
- Enforcing formatting or safety constraints
Mechanisms though, vary. Syntax, strictness and length allowed can vary widely. As of 2026, some like Claude and Gemini allow for very long instruction blocks.
Instructions in Studio AI
When the AI interprets a new prompt from you, it takes into account the instructional prompting that is sent along with it.
With the introduction of AI-Chat in the Valentina Studio 16.5 SQL Editor, not only can you leverage all major LLMs, but you can designate a set of instructions to send along with your prompts.
How to Set Instructions
- With AI-Chat visible, click the Preferences…button to display the Chat Settings… Dialog.
- Click on the model you want to use to reveal two tabs: Settings and Instructions.
- Click on the Instructions Tab. Type in or paste your instructions into the Instructions Field.
When you close the Chat Settings Dialog, your instructions are automatically saved.
What Happens Next?
Each time you send a prompt to the selected LLM, the LLM also receives the instructions, and employs those instructions when it decides what to reply with, when it responds.