⬜ AI text request

The Text AI Query block allows you to connect a scenario to text-based AI models (LLM) to generate dynamic and contextual responses, providing fluid and personalized interactivity.

This block analyzes the conversation history, the role assigned to the AI and the user prompt to produce a new response, adapted to the context. It eliminates the limitations of predefined responses by allowing free and scalable exchanges.

📥 Entries

• Prompt (Text): The instruction or question provided by the user (example: "How do you make a potion?").
• System prompt (Text - Optional): The personality or behavior rules that the AI must respect (example: "You are a wise old sarcastic wizard who refuses to help without flattery").
• Model (Dropdown): The AI model used to generate the response. Available values: mistralai/ministral-3b-2512 (default value) and google/gemini-2.5-flash.
• Temperature (Number, default value: 1): Adjusts the randomness/creativity of the generated response.

📤 Outputs

• Result (Text): The text produced by the AI in response to the prompt and the context (example: "A potion? Sigh. Another amateur. Show a little enthusiasm, and we'll see.").
• Next (Stream): Signal triggered once the response has been generated. The block has a single flow output (no dedicated error output): if the request fails, the flow still continues with an empty result.

💡 Example of use

Scenario: Dialogue with a traveling merchant

1. The player interacts with a character via a Text entry block: "What are you offering for sale?".
2. The entered text is passed to the AI Text Query block via the Prompt input.
3. The System prompt is defined as: "Clever merchant who exaggerates the value of his items and responds in rhymes".
4. The block generates a response after a few seconds of processing.
5. On Next, the Result is displayed: "Rare treasures, my friend! A sword that shines like day, or a potion that cures love!".
6. The flow continues based on the player's response (purchase, negotiation, etc.).

⚙️ Technical Details

• Token management: History consumes tokens (AI memory). To optimize resources, limit the history to recent and relevant exchanges.
• Latency and moderation: Errors can occur in the event of a response time that is too long or content blocked by moderation filters.

Documented version: v1

Updated on: 06/09/2026