Enable [[Project:Terminology gadget|Terminology gadget]] by default?
That's in fact omore complicae than that in various languages, which also require special aggutination (no prefixes, suffixes, infixes), mutations or reductions (e.g. transforming "de le" into "du" in French, though that those term would not be in the glossary), contextual elision (frequent in French or Italian for pronouns, articles, and particles), deagglutination (think about German verbs whose prefix can be detached and moved to the ned of sentence as in: "(Something) angehen" - infinitive; into: "Ich gehe (something) an." - indicative). In French and English we can also have a variation about whever some words or prefixes should be glued, attached by an hyphen, or detached, As well there are wellknown words with several accepted orthographies (depending on whever an ortographic reform was applied or not).
Now consider the case of conjugating verbs; there are too many forms (and listing them exhaustively requires lot of data: see for example conjugation rules in French Wiktionnary; and they are still not exhaustively listed, because feminine or plural variations of their participles depending on other words are not shown). Complex rules also exist for Slavic languages, and usually verbs are listed by their infinitive (but this is not the case for Latin where traditioanlly they are listed by their first person present indicative, or by the two first persons and the infinitive; but there are also many defective forms where only some persons or tense mode are existing, so this is not a general rule...).
If we start enumerating all forms, there's no well defined order. In my opinion we should just list the main lemma entry found in dictionnaries, e.g. within Wiktionnary, which could be linked by a lexical Wikidata item, that would list all their other forms (but attaching links to other languages will remain very difficult as it is hard to bind lemmas, where it is simpler to bind specific forms to their main lemma entry within the same language (but there are also complex exceptions there).
Attaching lexemes is a good idea, at least for now we may even just link these in the notes section, I had not thought of this.
What I had in mind though was that in including a limited selection of forms rather than a full conjugation table, it would be helpful to parse that grammatical information more easily. The nature of translating for software interfaces necessarily limits the contexts in which different words are used, making particular forms more used than others. For example, a typical Punjabi verb has over 100 possible forms. However, for some we may only need imperatives and for others we may only need imperfect participles. There are also some considerations for underlying meaning, for example, for the verb 'to put' I am only using the politest imperative form پایو rather than پاؤ which sounds forceful or پا which is just rude. So I would indicate that this particular form is the pertinent one, along with four perfect participle conjugations and a conjunctive participle form. For other verbs though, like 'to search,' I would make it more polite if the interface is telling the user to search as کھوجیو but normal politeness for the user telling the software to search as کھوجو. (Even though the software is not a person, being rude or informal with the software would not be read well.)