Schema appears to have two plural forms that are both valid: schemata and schemas. Are they completely interchangeable; or are there any guidelines on which one is appropriate for particular contexts?
The Greek schemata is mostly found in academic writing (e.g. experimental schemata), whereas the anglicized schemas is used freely in both technical and more general discourse (e.g. database access schemas).
When a word comes into English from another language (or as a coinage using parts from another language) with a countable noun sense, there are two possible approaches to pluralising; to also borrow the plural (hence criteria for criterion, and bacteria for bacterium) or using the normal -s and similar productions of English plurals (pianos or the longer pianofortes rather than the Italian piani and pianoforti).
There's no logical rule that says we can't say "The musicians got sick because the piani were covered with bacteriums", but language exists in a space of rough consensus, and the consensus is that piani and bacteriums are wrong, while pianos and bacteria are correct, and we keep using them because that's what everyone else does.
While many words go strongly one way or another, there are several words in English that have both a loan-word plural from the same source as the singular and one produced in the normal -s production for English plurals: radius with radii and radiuses, cactus with cacti and cactuses, virtuoso with virtuosi and virtuosos.*
As with things like which spelling became the accepted spelling, whether a coinage or slang word becomes considered "a real world" and so on, it's hard to predict which will happen until a point of consensus is reached at which point one form becomes "just right" and another "just wrong", or alternatively two forms continue to compete for a long time (consider lit and lighted competing for 400 years to be the past of light as an example of the same thing happening with a different case where there can be a comparable split).
There are a few things that can encourage it to go one way or another. The more the average English-speaker who uses the word is to have at least a passing knowledge of the source language, the more likely it is to survive in English (hence Latin and Greek plurals survive more due to their former educational importance). A language that doesn't often have plural forms will often not have that lack of plural form continue into English (hence "two ninjas" is more likely than "two ninja"). Words that are commonly used in the plural will more often keep them (hence criteria and bacteria survive easily) or even overthrow the singular (data is often used as an uncountable or even singular, displacing datum).
Contrasting to that, if a word is relatively obscure, then so will its plural be, and people will produce an -s plural lacking any other choice.
As a general rule, usages will become more and more "normal" over time, and that includes words with foreign plurals tending to become -s plurals.
This ngram would suggest that this has been happening with schemas/schemata: Bearing the fallibility of ngrams in mind, it would still seem that schemata was the more popular for some time, but schemas overtook recently.
Since people use language as those around them use it, use in a particular community may go strongly one way or another, and it could be that this is at play here. The term schema comes up often in terms of databases today, and many people using the word in that context would be otherwise unfamiliar with the word (and hence more likely to use a productive than an imported plural). Considering the following, they may be the community that has pushed schemas past schemata:
At the same time though, part of the SQL standard is called "SQL/Schemata" and some databases have a view or table called "Schemata" (a table in MySQL and a view in SQL Server and PostgreSQL), which is likely to keep it at least known about in that community.
The opposite pressure used to exist in the XML Schema standard as its documentation seemed to favour schemata once, though that community seems to have moved to schemas. Still, both exist in current use, including this announcement that uses one in one sentence, and the other in the next!
Are they completely interchangeable;
All the above is a way of arriving at "yes, they are completely interchangeable". While one of the other might be more common in certain contexts, there are none where one is so massively more popular as to make the other seem wrong, or even strange.
or are there any guidelines on which one is appropriate for particular contexts?
Not really, for that reason. If you're worried then use what others around you are using, but if you just like the look and sound of one more (I personally prefer schemata) then by all means do as you will.
*There are two other possibilities. One is that people may apply the wrong foreign plural rule, e.g. producing octopi. Another is that joke plurals may force foreign forms onto plurals that didn't generally use them, and later other people may not get the joke, leading some people to think that virii and penii are alternatives to, or even more "learned" than, viruses and penises when those were just jokes.