hotaru 0.8.0

Small, sweet, easy framework for full-stack web application
Documentation

Hotaru Web Framework

Latest Version Crates.io GPL-3.0 License

Small, sweet, easy framework for full-stack Rust web applications

License note: currently GPL-3.0; we are considering a change to LGPL or MIT (not decided).

Former Codebase

We rebased our code since July this year. Please refer to the following link if you are interested in our history of building the framework

https://github.com/Redstone-D/starberry

📋 Overview

Hotaru is a lightweight, intuitive web framework focused on simplicity and productivity. It supports regex-based routing, tree-structured URLs, and integrates seamlessly with the Akari templating system.

Example Project

MSRV: 1.86

✨ Key Features

  • Simple API: Intuitive request/response handling with minimal boilerplate
  • Full-Stack: Built-in template rendering with Akari templates
  • Flexible Routing: Support for regex patterns, literal URLs, and nested routes
  • Asynchronous: Built with Tokio for efficient async handling
  • Form Handling: Easy processing of form data and file uploads
  • Middleware Support: Create reusable request processing chains
  • Multi-Protocol Support: Handle HTTP/HTTPS, WebSocket, and custom TCP protocols
  • Security: Built-in request validation, size limits, and safety controls
  • Client Information: Access client IP addresses and connection details directly from handlers

🚀 Quick Start

use hotaru::prelude::*;

pub static APP: SServer = Lazy::new(|| {
    Server::new()
        .binding("127.0.0.1:3000")
        .build()
});

#[tokio::main]
async fn main() {
    APP.clone().run().await;
}

endpoint! {
    APP.url("/"),
    pub index<HTTP> {
        text_response("Hello, Hotaru!")
    }
}

📦 Installation

Using the CLI Tool (Recommended)

Install the Hotaru CLI tool:

cargo install hotaru

Create a new project:

hotaru new my_app
cd my_app
cargo run

Manual Installation

Add to your Cargo.toml:

[dependencies]
hotaru = "0.8.0"
tokio = { version = "1", features = ["full"] }

Optional Features

Hotaru supports the following optional features:

  • debug: Enable debug logging for development and troubleshooting

  • https: HTTPS support — pulls in hotaru_tls and surfaces HTTPS / TlsTransport / TlsOutboundTarget / TlsClientConfig

  • external-ctor: Use the external ctor crate instead of Hotaru's built-in constructor implementation

    Note: When enabling external-ctor, you must also add ctor to your dependencies:

    [dependencies]
hotaru = { version = "0.8.0", features = ["external-ctor"] }
ctor = "0.4.0"  # Required when external-ctor feature is enabled
tokio = { version = "1", features = ["full"] }

    By default, Hotaru uses a built-in constructor implementation that doesn't require any external dependencies.

🛠️ Binary Commands

Hotaru provides a CLI tool to help you scaffold and manage projects quickly.

⚠️ IMPORTANT: It is crucial to use the CLI tool (hotaru new or hotaru init) when creating a new Hotaru project. The CLI automatically generates essential files including:

  • build.rs - Required for asset management and resource copying
  • resource.rs - Helper module for locating templates and static files at runtime

Without these files, your project will not be able to properly locate and serve templates or static assets. Manual setup is significantly more complex and error-prone.

Installation

cargo install hotaru

Note: After installation, make sure Cargo's bin directory is in your PATH. If the hotaru command is not found, add Cargo's bin directory to your PATH:

# Linux/macOS
export PATH="$HOME/.cargo/bin:$PATH"

# Windows (PowerShell)
$env:Path += ";$env:USERPROFILE\.cargo\bin"

To make this permanent, add the export line to your shell configuration file (~/.bashrc, ~/.zshrc, etc.).

Available Commands

hotaru new <project_name>

Create a new Hotaru project with a complete project structure:

hotaru new my_app

This generates:

  • src/main.rs - Main application entry point with a hello world endpoint
  • src/resource.rs - Resource file locator helper module
  • build.rs - Build script for asset management
  • Cargo.toml - Pre-configured with Hotaru dependencies
  • templates/ - Directory for Akari HTML templates
  • programfiles/ - Directory for static assets (CSS, JS, images)

The generated project uses the latest Hotaru features including:

  • LServer! macro for clean server initialization
  • endpoint! macro for routing
  • Proper resource management with build-time asset copying

hotaru init

Initialize an existing Cargo project with Hotaru scaffolding:

cd my_existing_project
hotaru init

This command adds the same structure as hotaru new but to your current project directory.

Running Your Project

After creating a project:

cd my_app
cargo run

Your server will start at http://127.0.0.1:3000 by default, serving a "Hello, world!" response at the root path.

Project Structure

my_app/
├── Cargo.toml              # Dependencies and project metadata
├── build.rs                # Asset copying build script
├── src/
│   ├── main.rs            # Application entry point
│   └── resource.rs        # Resource locator helpers
├── templates/             # Akari HTML templates
└── programfiles/          # Static assets (CSS, JS, images)

The build script automatically copies templates/ and programfiles/ to the target directory during compilation, making them accessible to your application at runtime.

🎯 Core Concepts

Endpoints

Define routes with the endpoint! macro:

endpoint! {
    APP.url("/users/<int:id>"),
    pub get_user<HTTP> {
        let user_id = req.param("id");
        json_response(json!({ "id": user_id }))
    }
}

Macro Notes

  • endpoint! and middleware! auto-register at startup (constructor-based), so there is no manual router.register() step.
  • Always use brace syntax {} and place doc comments inside the macro block.
  • Optional fn-style: pub fn name(req: HTTP) { ... } (new in 0.7.7); angle-bracket form (hotaru blocks) defaults to req.
  • Our philosophy is to wrap anything into macros to keep endpoints and middleware self-contained; see macro_ra.md for the minimal syntax and rationale.
  • Analyzer support is planned via custom analyzer tools.

Middleware

Create reusable middleware:

use htmstd::session::CookieSession;

pub static APP: SServer = Lazy::new(|| {
    Server::new()
        .binding("127.0.0.1:3000")
        .append_middleware::<CookieSession>()
        .build()
});

Templates

Render HTML with Akari:

endpoint! {
    APP.url("/profile"),
    pub profile<HTTP> {
        let data = json!({ "name": "Alice" });
        template_response("profile.html", data)
    }
}

HTTP Safety Configuration

Configure request validation per endpoint:

endpoint! {
    APP.url("/upload"),
    config = [HttpSafety::new()
        .with_max_body_size(50 * 1024 * 1024)  // 50MB
        .with_allowed_methods(vec![HttpMethod::POST])
    ],
    pub upload<HTTP> {
        // Handle file upload
    }
}

Client Information

Access client IP addresses and connection details directly from your handlers:

endpoint! {
    APP.url("/api/whoami"),
    pub whoami<HTTP> {
        // Get client's full socket address (IP + port)
        match req.client_ip() {
            Some(addr) => text_response(format!("Your address: {}", addr)),
            None => text_response("Unknown client"),
        }
    }
}

Available Methods:

Method Return Type Description
client_ip() Option<SocketAddr> Client's socket address (IP + port)
client_ip_or_default() SocketAddr Returns 0.0.0.0:0 if unknown
client_ip_only() Option<IpAddr> Just the IP address, no port
client_ip_only_or_default() IpAddr Returns 0.0.0.0 if unknown
server_addr() Option<SocketAddr> Server's bound address
remote_addr() Option<SocketAddr> Alias for client_ip()
local_addr() Option<SocketAddr> Alias for server_addr()

Note: When behind a reverse proxy, client_ip() returns the proxy's address. Use headers like X-Forwarded-For or X-Real-IP to get the original client IP.

📚 Examples

Check out the example repository for:

  • Basic routing and handlers
  • Form processing and file uploads
  • Session management with cookies
  • CORS configuration
  • Multi-protocol applications

🔧 Crate Ecosystem

Hotaru is built on a modular architecture:

  • hotaru - Main framework with convenient API
  • hotaru_core - Core protocol and routing engine
  • hotaru_trans - Procedural macros for endpoint! and middleware!
  • hotaru_http - HTTP implementation for Hotaru
  • hotaru_tls - TLS/HTTPS implementation for Hotaru
  • hotaru_lib - Utility functions (compression, encoding, etc.)
  • htmstd - Standard middleware library (CORS, sessions)

📋 Changelog

0.8.0 (Current)

  • Client / outpoint runtime: new Client<TS> for outbound traffic, mirroring Server<TS>. Includes Client::request_fn, Client::call_fn, Client::call_url for one-shot and persistent outpoint invocation.
  • outpoint! macro: client-side counterpart to endpoint!. The user body becomes the outermost middleware; the send; marker triggers the registered chain (terminating in <P as Protocol>::send(ctx).await).
  • run! and call! macros: invocation-style sugar — run!(APP<HTTP>::name, request) -> APP.request_fn::<HTTP>("name", request); call!(APP<HTTP>::name) -> APP.call_fn::<HTTP>("name") (plus : "/path" form for call!).
  • Protocol trait reshape: channel-based open_channel / handle(channel, runtime, root) / send(ctx) -> ctx, plus a per-protocol install_channel(&mut ctx, channel) bridge. type Context: RequestContext<Channel = Self::Channel> pins channel-type alignment.
  • RequestContext trait: Default supertrait; declares type Channel: Channel as a type anchor only (channel lives as a private field on each concrete context — no trait-exposed accessor). Adds inject_request / into_response. type Error: ProtocolError is the single source of truth for chain errors.
  • Result-typed execution chain: middleware, final handlers, ExecutionChain::run, and UrlNode::run all return Result<C, <C as RequestContext>::Error> — no boxing at the chain boundary.
  • Named access points: every registered endpoint/outpoint has an explicit name. ProtocolEntry::register(name, path, step_names, binding, config) is the single canonical funnel; AccessPointTable per protocol entry refreshes Node-variant entries on rebind. Server::url(url, name, ...) / Client::url(url, name, ...) are URL-first.
  • Instance-based transports: TransportSpec::Inbound / Outbound replace the old Accepter / Connector shape. TcpInbound::bind(target) and TcpOutbound::build(target) materialize once per Server/Client.
  • HTTPS feature: new https cargo feature on the umbrella hotaru crate forwards to hotaru_http/tls, surfacing HTTPS, TlsTransport, TlsOutboundTarget, TlsClientConfig. HTTPS = Http1Protocol<TlsStream, TlsTransport>.
  • Channel trait + ProtocolFlow: per-exchange close semantics (HTTP/1 closes the connection; multiplexed protocols may close only the stream). Channel: Clone is the framework's sharing contract; protocols are free to back the handle with Arc<...> internally.
  • New LServer!, LClient!, LUrl!, LPattern! macros for simplified lazy static declarations (replaces the old LApp!). Defaults to TcpTransport; HTTPS clients with Client<TlsTransport> currently need a direct Lazy::new(...) declaration until the LClient! type-parameter passthrough lands.

0.7.x

  • Multi-protocol support (HTTP, WebSocket, custom TCP)
  • Enhanced security controls with HttpSafety
  • Improved middleware system with protocol inheritance
  • Performance optimizations in URL routing
  • Comprehensive security testing
  • .worker() method now properly configures dedicated worker threads per Server instance
  • Fixed hotaru new and hotaru init to generate correct endpoint! macro syntax
  • Built-in constructor implementation (no external ctor dependency required)
  • Fn-style blocks: New syntax pub fn name(req: HTTP) { ... } for endpoint! and middleware! macros (original hotaru blocks syntax preserved)
  • Bug fix for URL routing

0.6.x

  • Protocol abstraction layer
  • Request context improvements
  • Standard middleware library (htmstd)
  • Cookie-based session management

0.4.x and earlier

  • Async/await support with Tokio
  • Akari templating integration
  • Cookie manipulation APIs
  • File upload handling
  • Form data processing improvements

📚 Learn More

Video Resources URL
Quick Tutorial Youtube: https://www.youtube.com/watch?v=8pV-o04GuKk&t=6s Bilibili: https://www.bilibili.com/video/BV1BamFB7E8n/

🤖 AI Declaration of each Mod

We believe in transparency about AI-assisted development. The framework is governed jointly by two maintainer groups using a shared four-tier system that prioritizes understanding over line counts.

Maintained by: PMINE/Research

Name Tier Comments
hotaru_core/app Author-Owned
hotaru_core/connection Author-Owned
hotaru_core/executable Author-Owned
hotaru_core/url Author-Owned
hotaru_core/protocol Author-Owned
hotaru_http/trails Co-Authored
hotaru_http/* Human-Led
hotaru_mqtt/broker Co-Authored
hotaru_mqtt/traits Co-Authored
hotaru_mqtt/* Human-Led
hotaru_lib Human-Led Basic API Access
h2per Co-Authored Integration of Hyper - Not stable yet
htmstd/cors Human-Led
htmstd/session Human-Led

Maintained by: Project-StarFall

Name Tier Comments
hotaru_trans/endpoint Author-Owned Proof and language design must be fully understood by humans
hotaru_trans/outpoint Author-Owned Proof and language design must be fully understood by humans
hotaru_trans/middleware Author-Owned Proof and language design must be fully understood by humans
hotaru_trans/cors Co-Authored Trivial user-level abstraction
ahttpm Co-Authored Imports akari_macro plus improvements
SFX Co-Authored Trivial user-level abstraction
akari External https://crates.io/crates/akari
akari_lang External (TBD)
akari_macro External https://crates.io/crates/akari

Shared term meanings

Term Meaning
Forbidden The intelligence work in this module — design decisions, proof obligations, language semantics, novel logic — is authored by humans. AI is not used for this content (the mechanical/test/doc carve-out in operating rule 2 still applies). Reserved for modules where the work is the thinking, not the typing.
Author-Owned AI may assist with drafts and completion, but the committed code reads as the author's own throughout. A reviewer should not be able to tell where AI helped. The module signals "a human owns the design and the prose."
Human-Led The human authored the structure and the load-bearing pieces; AI filled in helpers, repetitive sections, or boilerplate. Some sections may visibly bear AI's hand, but the design choices and non-trivial logic are clearly human. The author can defend every part without re-consulting AI.
Co-Authored AI participated substantively in both design exploration and implementation. The human author has internalized the result and can defend, modify, and debug without re-prompting. Appropriate for well-understood patterns and third-party integrations.

The understanding requirement is uniform across Author-Owned, Human-Led, and Co-Authored: the author can explain any line, modify surrounding code without AI help, and walk a reviewer through the code on request. The tiers differ only in where AI's voice is allowed to show through, not in what the author owes the team.

Operating rules

  1. No quantification. Tiers describe the kind of collaboration, not the amount of AI-authored code. Counting lines is brittle and incentivizes the wrong behavior.
  2. Tests, documentation, and mechanical typing are permitted in every tier — including Forbidden. AI assistance is allowed across all modules for: unit tests; doc comments and inline prose; and mechanical typing where the design decision has already been made by a human and AI is only writing it out (e.g., applying a settled pattern across similar cases, expanding hand-authored pseudocode, regenerating a table from a spec). The defining criterion is that no intelligence work is delegated to AI — design, proof, and semantics decisions remain with the human. The author remains responsible for understanding what was generated.
  3. Reviewer-driven understanding check. Any reviewer may flag a PR with "this doesn't feel author-owned" — regardless of the module's tier. The author clears the flag by demonstrating understanding in PR comments or a short walkthrough. Flags are requests for evidence, not accusations.
  4. Smell-test threshold scales with tier. Author-Owned code is flagged if any section visibly reads as AI-generated. Human-Led is flagged if the structural code reads as AI-generated or AI's hand pervades rather than appearing locally. Co-Authored is flagged only if the author cannot defend the code in review.
  5. Tier reflects the work, not preferences. Maintainers set tiers based on the nature of the module. If the character of a module changes, the tier is re-set rather than stretched.
  6. External code is outside this policy. AI-authored code arriving through a third-party crate is governed by that crate's own conventions, transparently linked.

📄 License

GPL-3.0 License

Copyright (c) 2024-2026 @ Field of Dreams Studio

Project-StarFall, PMINE