Typed Java boundary adapter generation: deferred
Decision
This integration milestone does not generate Java-to-Sprig or Sprig-to-Java boundary adapters. The Fabric starter keeps a small, ordinary Java bridge and uses Sprig conform for the application-owned implementation.
Existing tools
sprig apiinspects Java members and reports which signatures the compiler can use, including the reason a signature is unsupported.sprig wrapgenerates editable Sprig source for supported Java members. It is a Java-to-Sprig wrapper generator, not a generator for Java classes that adapt Sprig implementations to arbitrary host APIs.conformchecks a bounded foreign Java-interface contract for an existing Sprig class. It does not synthesize methods or change the class ABI.
These tools answer useful inspection, import and conformance questions, but they do not remove the repeated Gradle bridge compilation and classpath wiring observed in the Quest Board experiment. sprig-gradle addresses that build mechanics directly while leaving the host boundary visible and reviewable.
Why generation is deferred
The Quest Board has one concrete bridge shape. That is enough to justify first-party Gradle integration, but not enough evidence to define a general adapter API. A generator would need independently useful fixtures for multiple real boundary shapes, explicit rejection rules for unsupported signatures, ordinary reviewable generated source, no runtime reflection, and a stable visible ABI. No new language semantics are authorized by this milestone.
Reconsider only after at least two independent host integrations show the same adapter boilerplate remains after using sprig-gradle. Start with a small design proposal that compares handwritten bridge source with existing api/wrap/conform contracts and identifies signatures that can be safely generated. Keep unsupported overloads, wildcards, callbacks, nullability and exception behavior explicit; never make a generator silently broaden the JVM interop contract.
