Crate swarm_engine 

SwarmEngine

High-throughput, low-latency Agent Swarm orchestration framework.

SwarmEngine provides a framework for orchestrating multiple AI agents working together on complex tasks. It supports various LLM backends and includes built-in evaluation capabilities.

Quick Start

use swarm_engine::prelude::*;

// Create an environment and orchestrator
let config = SwarmConfig::default();
let orchestrator = OrchestratorBuilder::new(config)
    .environment(env)
    .manager(manager)
    .build()?;

// Run the swarm
orchestrator.run().await;

Feature Flags

• eval - Enable evaluation framework
• ollama - Enable Ollama backend
• llama-server - Enable llama.cpp server backend
• llama-cpp - Enable llama.cpp native backend
• cuda - Enable CUDA support
• metal - Enable Metal support (Apple Silicon)
• full - Enable eval + all HTTP-based LLM backends

Re-exports

pub use swarm_engine_llm as llm;

Modules

actions

Actions 統一管理

agent

SwarmEngine Agent 定義

analysis

Analysis System - 状況分析

async_task

SwarmEngine AsyncTaskSystem

config

設定・パス管理モジュール

context

Context System - スコープ付きコンテキスト管理

debug

Debug - Tick単位のデバッグ出力機構

environment

Environment - アクション実行環境の抽象化

error

SwarmEngine 統一エラー型

events

Events - 行動イベントの定義と配信

exploration

Exploration System - 探索アルゴリズム (V2)

extensions

Extensions - 動的 Resource 管理

learn

Learn Module - 学習系 Domain の分離

online_stats

Online Stats - オンライン統計の提供

orchestrator

SwarmEngine Orchestrator

pipeline

Pipeline abstraction for event-driven processing.

prelude

Prelude module for convenient imports

snapshot

Snapshot Module - デバッグ・分析用のスナップショット機能

state

SwarmEngine State - 2層メモリモデル

telemetry

Telemetry - Swarm Engine 観測層

types

SwarmEngine 基本型定義

util

Utility modules