Let's take a closer look at the error messages you received:
When trying to call TokenService.getToken(...).then with a function bound to onFulfill, Flow indicates that property read is missing in String [1] within the first argument.
This error message means that Flow believes that result is a String, and therefore cannot access the property result.read as Strings do not have this property ("property read is missing"). Instead of focusing on logging the variable type of read, it would be beneficial to log its contents and verify if it is indeed a valid String.
The second error message states:
Attempting to call download with read bound to readToken reveals that the object type [1] is incompatible with string [2].
This implies that Flow perceives the type of read as something other than a String in the line:
download(url_template, name, read, is_watermarked);
It is important to ascertain whether read is genuinely a string before attempting to set its type or override it (in Typescript, this would typically be done through (read as string), although I am uncertain about how Flow handles this situation).
Remember that what you observe in the console at runtime differs from Flow's assumptions about a variable's type, as Flow can only make educated guesses regarding a variable's type without certainty about its behavior during runtime.
===========================
I want to clarify that the following answer is inaccurate; it has been retained here to avoid confusing individuals who may come across comments referencing this response
The issue arises when examining the result object in this context:
const { read } = result;
If you notice that the type of result is not explicitly defined, you can enforce a strong type for the read variable under the assumption that your API consistently delivers correct responses with strings located under result.read. This can be accomplished by casting the read variable as a string:
download(url_template, name, (read as string), is_watermarked);
Please bear in mind that inspecting a variable through logging differs from working within a Typescript environment. Your approach essentially verified the runtime value, whereas Typescript primarily operates at the build level and might not anticipate the precise nature of the API responses.