If the message text appears to be valid, the function returns non-zero.
If the message text is empty, contains invalid characters, or appears to
contain non-text data, the function returns zero. To get extended error
information, call LlmGetLastError.
The LlmValidateText function is intended to help applications
detect invalid or malformed input before sending a request to a model.
It checks for common problems such as empty strings, text that consists
only of spaces or line breaks, illegal Unicode code points, non-printable
characters, and input that appears to contain binary data. The message
text cannot contain embedded null characters.
Applications that accept messages from users or other external
sources should validate the text before submitting it to a model. This can
help detect malformed or invalid input before a request is sent, reducing
the chance of unexpected results or incurring unnecessary usage costs.
This function does not evaluate the quality, safety, or semantic
meaning of the message. Although this function is most commonly used
to check prompts sent to a model, it can be used to validate any type
of text to ensure it is not malformed and contains printable
characters.
There are two versions of this function. The wide character version of
this function (LlmValidateTextW) validates the message text as UTF-16
Unicode. The ANSI version (LlmValidateTextA) first attempts to
interpret the input as UTF-8 encoded text. If the input is not valid UTF-8,
it attempts to convert the text to Unicode using the active
code page and the resulting Unicode text is validated. The generic function
name LlmValidateText automatically resolves to the ANSI or Unicode
version based on whether the application is compiled with Unicode support.