apache/opendal

Apache OpenDAL solves a surprisingly common headache: every storage service, AWS S3, Google Cloud Storage, Azure Blob, local files, FTP servers, Dropbox, Hugging Face, has its own API with its own quirks. If your app needs to read from or write to several of these, you normally write and maintain separate integration code for each one. OpenDAL gives you a single, unified interface so you can swap or combine storage backends without rewriting your application logic. The way it works is straightforward in concept. You pick the storage services you need, say, S3 for cloud object storage and a local filesystem for caching, and talk to them through the same API. On top of that, you can layer in cross-cutting concerns like retry logic, timeouts, logging, tracing, metrics, and concurrency limits. These layers apply regardless of which storage backend you're talking to, so you configure them once instead of implementing them per service. The core is written in Rust, and there are bindings for Python, Java, Go, Node.js, Ruby, C, C++, Swift, Dart, and several other languages, so you can use it from whatever stack you're already in. This is useful for anyone building data-heavy applications. A startup storing user uploads might start with local disk and later move to S3 or Google Cloud Storage without touching application code. A data team pulling model artifacts from Hugging Face and writing results to Azure Blob can use one interface for both. A product team building a backup tool that supports Dropbox, Google Drive, and OneDrive can integrate all three through the same abstraction. What's notable is the breadth: the project supports dozens of services across object storage, file systems, cloud SaaS platforms, databases, standard protocols like FTP and WebDAV, and key-value stores. The composable layer design means you only enable what you need, keeping things lean. The Rust core is designed to be zero-cost, meaning the abstraction itself shouldn't add meaningful overhead compared to talking to each service directly.