echo

A very fast distributed replay buffer for reinforcement learning. The core is a lockfree, pre-allocated Rust ring buffer; batches come back to Python as zero-copy numpy views, with the GIL released while you wait.

What it does

Distributed RL training pipelines typically have many hundreds or even thousands of rollout workers producing samples and a small number of learners consuming batches of them. echo exists because data transfer and stacking is often the bottleneck of these systems. It receives individual samples, assembles them into fixed-size batches, and serves those batches with minimal copying and no Python-side contention.

The core is a lockfree ring buffer written in Rust. Clients push samples into SPSC queues over the network, a pool of drainers moves data into the ring buffer, and the consumer pulls full batches out. Pytrees are flattened with optree on the client and unflattened on the Python server. Rust itself is pytree-agnostic.

Features

• Lockfree ring buffer, pre-allocated (no copies after ingress)
• TCP transport on the wire (custom transports — gRPC, RDMA, … — can be plugged into the same Rust Transport trait)
• Pytree-shaped samples (nested dicts/tuples of numpy arrays) via optree
• Zero-copy batches: sample() returns numpy views into Rust-owned memory
• GIL released while waiting for batches
• FIFO sampling
• Detailed metrics exposed per batch

Where to start

• Getting started: install, then a runnable example.
• Zero-copy batches: ownership and lifetime rules for the numpy views returned by sample().
• Python API: reference for every public class.
• Rust internals: how the lockfree path is put together.
• Reading metrics: glossary for every field on SampleInfo.