Interface SafetyFilter

public interface SafetyFilter
Pre-flight gate that inspects messages before they're sent to the model. Implementations may call a moderation API, run a local profanity match, or anything else. Returning a non-null reason blocks the chat call and propagates an LlmException whose LlmException.getType() is INVALID_REQUEST.

  • Field Summary

    Fields
    Modifier and Type
    Field
    Description
    static final SafetyFilter
    ALLOW_ALL
    A built-in filter that allows everything.

  • Method Summary

    Modifier and Type
    Method
    Description
    String
    check(List<ChatMessage> messages)
    Returns null to allow the call, or a human-readable reason string to block it.

  • Field Details

    • ALLOW_ALL

      static final SafetyFilter ALLOW_ALL
      A built-in filter that allows everything. Useful as a default or as a base for composition. Use SafetyFilters.openai(key) (in com.codename1.ai.filters or a separate cn1lib) for the OpenAI Moderation gate.

  • Method Details

    • check

      String check(List<ChatMessage> messages)
      Returns null to allow the call, or a human-readable reason string to block it.
      Parameters:
      messages - immutable request conversation
      Returns:
      null to allow submission, otherwise the rejection reason