Foundry 1.0.beta
=================

This is our beta for the upcoming GNOME 49 release.

I do anticipate additional API additions before release candidates.

 * A new feature flags system which allows compiling out major portions of
   libfoundry for situations where they are not necessary or desired. I would
   expect this mostly used for static linking situations as it does affect
   the ABI produced by the library.

 * Various new CLI commands such as `llm complete`, `test list`, `test run`,
   `clone`, `vcs log`.

 * The build and run subsystems have grown GActions and support for wriring
   up PTY for those actions.

 * Their is a new input abstraction which is used for template input,
   providing input to auth providers, and more in the future.

   A TTY input mechanism now exists on FoundryCommandLine to make reading
   input in an interactive manner simple.

 * A new word-completion provider which can scan input across included
   files similar to Vim.

 * New pre/post load operations for text document addins. Spellcheck has
   been rebuilt on top of this mechanism to simplify weak ref management.

   Document loading has been cleaned up significantly as part of this.

 * A JSONRPC subsystem has been implemented using a new JsonrpcDriver
   helper to support multiple jsonrpc dialects. As such, the dependency
   on jsonrpc-glib has been dropped.

 * A new test management subsystem has been added to extract tests from
   the active build system.

 * Improved support for code actions on text documents.

 * New CTags integration.

 * A gutter renderer for diagnostics has been added.

 * Improved bridge for Foundry based addins to GtkSourceView equivalents
   like hover, completion, and indenters.

 * A new symbol API has been implemented to provide introspection on
   symbols and their hierarchy.

 * Many version control abstractions have landed with an implementation
   for the Git backend.

 * The git backend has been rewritten using libgit2 directly and wrapping
   the git structures in a more sound way. Additionally threading has been
   used to help keep operations off the main thread when they have a
   potential for blocking IO.

 * A new LLM subsystem that can list models and request completions for
   said models.

 * A new Ollama plugin which provides scaffolding for applications to talk
   to a local model if they wish.