Hi everyone,
We’re excited to share that Lazy Dynamics has released a Python SDK for RxInferServer, making it easy to call high-performance RxInfer models from Python.
But there’s more:
RxInferServer itself is designed as a general-purpose backend that supports remote execution of probabilistic models, and with minor modifications, it can be used to call arbitrary Julia code, too. It also exposes an OpenAPI specification, making integration with other languages and frameworks straightforward.
What you can do with it today
- Use Python to call reactive probabilistic models hosted on RxInferServer
- Send/receive multidimensional arrays and distributions from JSON
- Run inference with RxInfer models remotely
- Use the OpenAPI interface for auto-generated clients or integration in other tools
- Extend the server to call custom Julia code, not just RxInfer models
- The server supports hot-reloading of code and models for fast and easy experimentation
Example + Docs
- Example (Python → RxInfer): Inference in State Space Models - RxInfer Client
- Full RxInferServer docs: Home · RxInferServer
- RxInfer website: https://rxinfer.com
What’s next?
- More examples: online inference, real-time data pipelines, deployment patterns
- Better Julia integration examples (beyond RxInfer)
- More SDKs for RxInfer in JavaScript/TypeScript/You name it
We’re looking for early feedback from developers interested in:
- Bayesian inference and probabilistic programming
- Bridging Python and Julia for ML workloads
- Real-time, reactive modeling
Feel free to try it out, report bugs, suggest features, or just tell us how it feels from your perspective as a dev.