Custom providers: storage, vectors, etc.

[natoverse]

With the v3 release we've significantly revamped and streamlined how underlying providers are managed. We have a factory pattern that we use throughout that allows you to register a custom implementation and use settings.yml to configure it. This covers language model providers, vector stores, backend storage, chunkers and input processors, and more. We'll be closing a lot of issues and PRs related to "can you add x?", because we just don't have the capacity to maintain more providers than the core ones we already support. But if you have a custom implementation folks can use, feel free to post links here.

More info on the factory pattern can be found here: https://github.com/microsoft/graphrag/blob/main/packages/graphrag-common/README.md

[aniruddhaadak80]

This factory pattern approach is solid. I appreciate that the design leaves room for community implementations instead of forcing everything into one maintained set of providers. What I would want clarified next is which provider gaps come up often enough that they may deserve first-class examples or support.

It would also help to document how custom providers are expected to behave across version changes. If someone builds against one release and the interfaces shift later, a little guidance on compatibility expectations would make the extension story feel much safer.