Skip to main content

Documentation Index

Fetch the complete documentation index at: https://watermelon.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

It can sometimes happen that your AI Agent suddenly replies in a different language than expected. While this might feel confusing, this behavior is completely normal for all Large Language Models (LLMs). These models continuously adapt based on context, input, and configuration.
Fortunately, this is easy to fix.

​
Why does an AI Agent switch language?

LLMs determine their response language based on:
  • The language of the conversation
  • The language of the available knowledge
  • System-level instructions and context fields
If some of these inputs are inconsistent (for example, mixed languages), the AI Agent may switch languages during a conversation.
This is expected behavior and not a bug.

​
How to fix it

To make sure your AI Agent consistently responds in the correct language, you need to set one specific field in English. Once this is done, the AI Agent will correctly interpret and apply all language instructions again.
If your Agent was created some time ago already and still has Domain knowledge, please check the information below for Legacy Agents.
  1. Open your Agent.
  2. Navigate to AI Agent settings.
  3. Translate the personality to English (one of the other fields works too).
  4. Save your changes.
Once updated, your AI Agent will immediately apply the correct language consistently.

​
For Legacy Agents

  1. Open your Agent.
  2. Navigate to Domain knowledge.
  3. Translate the personality to English (one of the other tabs works too).
  4. Save your changes.
After this, the AI Agent will correctly pick up the intended language behavior.
A Legacy Agent includes Domain knowledge. You can recognize a a Legacy Agent by opening the Agent and checking if the Domain knowlegde is visible in the left-hand menu.