1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
//! `openkind-engine`: Runtime-agnostic decision engine abstractions and dispatch.
//!
//! # Architecture & Responsibilities
//! `openkind-engine` defines the central execution abstractions for Jev-compatible decision inference.
//! Per `docs/ARCHITECTURE.md`, the engine layer sits between `openkind-core` and the transport
//! layers (`openkind-api`):
//!
//! `core` ⇐ `engine` ⇐ `api` ⇐ `server/cli`.
//!
//! The engine abstraction is completely transport-agnostic: it accepts a [`SystemRequest`] and returns
//! a [`SystemResponse`], remaining oblivious to whether evaluation was triggered over HTTP/REST or gRPC.
//!
//! # Core Components
//! - [`DecisionEngine`]: Trait implemented by inference backends (e.g. [`MockEngine`], Candle, GGUF/llama.cpp, ONNX).
//! - [`EngineRegistry`]: Thread-safe mapping of public model aliases (e.g. `"jev-latest"`, `"mock"`) to engine instances.
//! - [`dispatch`]: Unified entrypoint that orchestrates validation, metrics recording, token estimation, and backend evaluation.
/// Decision request dispatching, validation, and token accounting.
/// DecisionEngine trait and token estimation helpers.
/// Engine-specific error types and result aliases.
/// In-memory mock decision engine implementation for testing.
/// Model execution profile contracts and parity specifications.
/// Thread-safe registry mapping model aliases to decision engines.
pub use dispatch;
pub use DecisionEngine;
pub use ;
pub use MockEngine;
pub use ;
pub use EngineRegistry;