dokime-server

Responsibility: adapt typed native gRPC services to dokime-app.

Public Surface

The crate exposes grpc::Server, grpc::Config, grpc::TlsIdentity, and grpc::Error. The generated service adapters are grouped internally by capability and implement the services declared in proto/dokime/v1/.

The adapters handle runtime queries and commands, installed artifact configuration flows, suite and cycle workflows, telemetry/plot streaming, reports, documents, artifacts, diagnostics, and branded assets.

Boundary

The server does not own hot state, scheduling, or durability policy. Those remain in dokime-app.

Attached server lifecycle

Server::start(runtime.handle(), config).await binds a native gRPC listener and serves the existing runtime. address() returns the bound socket address and local_endpoint() returns its native endpoint URI. shutdown().await stops accepting calls and drains gRPC connections; after the graceful drain window it cancels only listener-owned connection I/O. It does not stop the runtime. The owning host shuts the runtime down after releasing all other handles.

The typed services are documented in the gRPC reference. There is no HTTP route, WebSocket protocol, OpenAPI layer, or separate asset server in this crate.

Native appearance

grpc::Server registers the generated Appearance.GetAppearance and Content.DownloadAppearanceAsset methods under Read authorization. Appearance returns the complete resolved theme and branding data, with portable digest-bound logo references. Downloads capture a registered image of at most 8 MiB before streaming metadata, bounded chunks, and a digest completion marker. They accept no caller filesystem paths. Refresh appearance after a missing or changed asset; clients verify both the digest and the final successful gRPC status.

The runtime remains the authority for brand precedence. Source strings and container entries are bounded before cloning, and encoded responses obey the configured message limit. These native methods require no backend UI.