Here's a minimal multilingual conlang database idea.

All words in all languages use the same lexeme table and the same pool of IDs. There is a language column to indicate which language a word belongs to. The words and meanings are stored in separate tables and connected via a senses table. If two words from the same language share a meaning they're synonyms. If two words from different languages share a meaning they can be translated back and forth.

The senses table has a rank column indicating how salient a particular meaning of a word is, with lower numbers being more salient. A sense with a rank of 1 is likely the meaning inferred by a listener without further context, and higher numbers refer to rarer senses of a word.

Importantly, lexemes and sememes can have zero or more entries in the senses table, meaning you can enter words or meanings on their own. Maybe you have a particular string of sounds but don't know what they mean yet, or have a particular concept you want to have a word for but don't know what that word is yet.