Crate modelexpress_server 

Re-exports

pub use server::run_server;

pub use cache::*;

pub use config::*;

pub use services::*;

Modules

auth

ServiceAccount TokenReview authentication plus an exact-match allowlist.

backend_config

Shared backend selection for the metadata stores (Redis, Kubernetes CRDs, and an in-memory backend behind the memory-backend feature for tests and local dev).

cache

config

metrics

Prometheus exposition for the ModelExpress server.

p2p

P2P model metadata: worker coordination, tensor descriptors, and NIXL RDMA transfer.

refit

RL refit control-plane service and metadata backends.

registry

Distributed model registry: model-download lifecycle (DOWNLOADING / DOWNLOADED / ERROR) and LRU cache-eviction timestamps.

server

Reusable server entrypoint. main is a thin shell over run_server so the whole startup path (registry, P2P state, health, reaper, graceful shutdown) can be embedded by a downstream binary that provides its own configuration or services.

services