Crate ares 

A.R.E.S - Agentic Retrieval Enhanced Server

A production-grade agentic chatbot server built in Rust with multi-provider LLM support, tool calling, RAG, MCP integration, and advanced research capabilities.

Overview

A.R.E.S can be used in two ways:

1. As a standalone server - Run the ares-server binary
2. As a library - Import components into your own Rust project

Quick Start (Library Usage)

Add to your Cargo.toml:

[dependencies]
ares-server = "0.5"

Basic Example

use ares::{Provider, LLMClient};

#[tokio::main]
async fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Create an Ollama provider
    let provider = Provider::Ollama {
        base_url: "http://localhost:11434".to_string(),
        model: "llama3.2:3b".to_string(),
    };

    // Create a client and generate a response
    let client = provider.create_client().await?;
    let response = client.generate("Hello, world!").await?;
    println!("{}", response);

    Ok(())
}

Using Tools

use ares::{ToolRegistry, tools::calculator::Calculator};
use std::sync::Arc;

let mut registry = ToolRegistry::new();
registry.register(Arc::new(Calculator));

// Tools can be used with LLM function calling
let tool_definitions = registry.definitions();

Multi-Turn Tool Calling with ToolCoordinator

use ares::llm::{Provider, ToolCoordinator, ToolCallingConfig};
use ares::tools::ToolRegistry;
use std::sync::Arc;

let provider = Provider::from_env()?;
let client = provider.create_client().await?;
let registry = Arc::new(ToolRegistry::new());

// Create a unified coordinator that works with any LLM provider
let coordinator = ToolCoordinator::new(client, registry, ToolCallingConfig::default());

// Execute a tool-calling conversation
let result = coordinator.execute(Some("You are a helpful assistant."), "What is 25 * 4?").await?;
println!("Response: {}", result.content);
println!("Tool calls made: {}", result.tool_calls.len());

Configuration-Driven Setup

use ares::{AresConfigManager, AgentRegistry, ProviderRegistry, ToolRegistry};
use std::sync::Arc;

// Load configuration from ares.toml
let config_manager = AresConfigManager::new("ares.toml")?;
let config = config_manager.config();

// Create registries from configuration
let provider_registry = Arc::new(ProviderRegistry::from_config(&config));
let tool_registry = Arc::new(ToolRegistry::with_config(&config));
let agent_registry = AgentRegistry::from_config(
    &config,
    provider_registry,
    tool_registry,
);

Feature Flags

Feature	Description
ollama	Ollama local inference (default)
openai	OpenAI API support
llamacpp	Direct GGUF model loading
postgres	PostgreSQL database (default)
qdrant	Qdrant vector database
mcp	Model Context Protocol support

Modules

• agents - Agent framework for multi-agent orchestration
• api - REST API handlers and routes
• auth - JWT authentication and middleware
• db - Database abstraction (PostgreSQL)
• llm - LLM client implementations
• tools - Tool definitions and registry
• workflows - Declarative workflow engine
• types - Common types and error handling

Architecture

A.R.E.S uses a hybrid configuration system:

• TOML (ares.toml): Infrastructure config (server, auth, providers)
• TOON (config/*.toon): Behavioral config (agents, models, tools)

Both support hot-reloading for zero-downtime configuration changes.

Re-exports

pub use agents::AgentRegistry;

pub use agents::AgentRegistryBuilder;

pub use db::tenants::TenantDb;postgres

pub use db::PostgresClient;postgres

pub use llm::client::LLMClientFactoryTrait;

pub use llm::ConfigBasedLLMFactory;

pub use llm::LLMClient;

pub use llm::LLMClientFactory;

pub use llm::LLMResponse;

pub use llm::Provider;

pub use llm::ProviderRegistry;

pub use models::ApiKey;

pub use models::Tenant;

pub use models::TenantContext;

pub use models::TenantQuota;

pub use models::TenantTier;

pub use tools::registry::ToolRegistry;

pub use types::AppError;

pub use types::ErrorCode;

pub use types::Result;

pub use utils::toml_config::AresConfig;

pub use utils::toml_config::AresConfigManager;

pub use utils::toon_config::DynamicConfigManager;

pub use workflows::WorkflowEngine;postgres

pub use workflows::WorkflowOutput;postgres

pub use workflows::WorkflowStep;postgres

Modules

agents

AI agent orchestration and management. AI agent orchestration and management.

apipostgres

HTTP API handlers and routes. HTTP API Handlers and Routes

authpostgres

JWT authentication and middleware. JWT Authentication and Middleware

cli

Command-line interface and scaffolding. CLI module for A.R.E.S

db

Database clients (Turso/SQLite, Qdrant). Database Clients and Vector Stores

llm

LLM provider clients and abstractions. LLM Provider Clients and Abstractions

mcpmcp

Model Context Protocol (MCP) server integration.

memory

Conversation memory and context management. Memory management module for conversation context and user memory.

middlewarepostgres

Middleware for API key auth and usage tracking.

models

Multi-tenant models (Tenant, ApiKey, Quota).

rag

Retrieval Augmented Generation (RAG) components. Retrieval Augmented Generation (RAG) Pipeline

researchpostgres

Multi-agent research coordination. Multi-Agent Research Coordination

tools

Built-in tools (calculator, web search). Built-in Tools for Agent Capabilities

types

Core types (requests, responses, errors). Core types used throughout the A.R.E.S server.

utils

Configuration utilities (TOML, TOON). Configuration utilities (TOML, TOON).

workflowspostgres

Workflow engine for agent orchestration. Workflow Engine Module

Structs

AppStatepostgres

Application state shared across handlers

Functions

base_routerpostgres

Returns the base ARES router with all generic endpoints.