axum

name	axum
description	Axum framework guardrails, patterns, and best practices for AI-assisted development. Use when working with Axum projects, or when the user mentions Axum framework. Provides Tower integration, async handlers, extractors, and type-safe routing guidelines.
license	MIT
metadata	{"author":"samuel","version":"1.0","category":"framework","language":"rust","extensions":".rs"}

Axum Framework Guide

Applies to: Axum 0.7+, Rust Web APIs, Microservices Complements: .claude/skills/rust-guide/SKILL.md

Core Principles

1. Tower-First: Axum is built on Tower; embrace Service, Layer, and middleware composition
2. Type-Safe Extraction: Use compile-time extractors for request data -- no manual parsing
3. Async Throughout: All handlers are async; never block the Tokio runtime
4. Modular Routing: Compose routers with merge and nest; separate public from protected routes
5. Structured Errors: Implement IntoResponse for custom error types; return appropriate HTTP status codes

Project Structure

myproject/
├── Cargo.toml
├── src/
│   ├── main.rs                # Entry point (thin: env, tracing, bind, serve)
│   ├── lib.rs                 # Module declarations
│   ├── config.rs              # Configuration loading (env-based)
│   ├── routes/
│   │   ├── mod.rs             # Router composition (create_router)
│   │   ├── users.rs           # User-related route definitions
│   │   └── health.rs          # Health check route
│   ├── handlers/
│   │   ├── mod.rs
│   │   └── users.rs           # Handler functions (thin: extract, call service, respond)
│   ├── models/
│   │   ├── mod.rs
│   │   └── user.rs            # Domain types, DTOs, validation
│   ├── services/
│   │   ├── mod.rs
│   │   └── user_service.rs    # Business logic layer
│   ├── repositories/
│   │   ├── mod.rs
│   │   └── user_repository.rs # Database access layer
│   ├── extractors/
│   │   ├── mod.rs
│   │   └── auth.rs            # Custom extractors (AuthUser, Claims)
│   ├── middleware/
│   │   ├── mod.rs
│   │   └── logging.rs         # Custom Tower middleware
│   └── errors/
│       ├── mod.rs
│       └── app_error.rs       # AppError enum + IntoResponse
├── tests/
│   └── integration_tests.rs
└── migrations/

Architectural rules:

• Handlers are thin: extract request data, call service, return response
• Services contain business logic; they call repositories
• Repositories handle database access only
• Models define domain types and DTOs with validation
• Errors are defined per-layer and composed with #[from]

Dependencies (Cargo.toml)

[dependencies]
# Web framework
axum = { version = "0.7", features = ["macros"] }
axum-extra = { version = "0.9", features = ["typed-header"] }
tokio = { version = "1.0", features = ["full"] }
tower = { version = "0.4", features = ["full"] }
tower-http = { version = "0.5", features = ["cors", "trace", "compression-gzip"] }

# Serialization
serde = { version = "1.0", features = ["derive"] }
serde_json = "1.0"

# Database
sqlx = { version = "0.7", features = ["runtime-tokio", "postgres", "uuid", "chrono"] }

# Validation
validator = { version = "0.16", features = ["derive"] }

# Error handling
thiserror = "1.0"
anyhow = "1.0"

# Logging
tracing = "0.1"
tracing-subscriber = { version = "0.3", features = ["env-filter"] }

# Configuration
config = "0.14"
dotenvy = "0.15"


 = { version = , features = [, ] }
 = { version = , features = [] }


 = 
 = 

Application Entry Point

// src/main.rs
#[tokio::main]
async fn main() -> anyhow::Result<()> {
    dotenvy::dotenv().ok();

    tracing_subscriber::registry()
        .with(tracing_subscriber::EnvFilter::try_from_default_env()
            .unwrap_or_else(|_| "myproject=debug,tower_http=debug".into()))
        .with(tracing_subscriber::fmt::layer())
        .init();

    let config = config::Config::load()?;
    let pool = sqlx::PgPool::connect(&config.database_url).await?;
    sqlx::migrate!("./migrations").run(&pool).await?;

    let state = AppState::new(pool, config.clone());
    let app = routes::create_router(state);

    let addr = SocketAddr::from(([0, 0, 0, 0], config.port));
    tracing::info!("Starting server on {}", addr);

    let listener = tokio::net::TcpListener::(addr).?;
    axum::(listener, app)
        .(())
        .?;

    (())
}

  () {
    tokio::signal::().
        .();
    tracing::info!();
}

Routing

Router Composition

// src/routes/mod.rs
pub fn create_router(state: AppState) -> Router {
    let cors = CorsLayer::new()
        .allow_origin(Any)
        .allow_methods(Any)
        .allow_headers(Any);

    let public_routes = Router::new()
        .route("/health", get(handlers::health::health_check))
        .route("/api/v1/auth/register", post(handlers::users::register))
        .route("/api/v1/auth/login", post(handlers::users::login));

    let protected_routes = Router::new()
        .route("/api/v1/users", get(handlers::users::list_users))
        .route("/api/v1/users/:id", get(handlers::users::get_user))
        .route("/api/v1/users/:id", put(handlers::users::update_user))
        .route("/api/v1/users/:id", delete(handlers::users::delete_user))
        .layer(middleware::from_fn_with_state(
            state.clone(),
            app_middleware::auth::auth_middleware,
        ));

    Router::()
        .(public_routes)
        .(protected_routes)
        .(TraceLayer::())
        .(CompressionLayer::())
        .(cors)
        .(state)
}

Routing Rules

• Use merge for flat route composition; use nest for path-prefix grouping
• Apply authentication middleware only to protected route groups
• Layer order matters: outermost layer runs first (CORS, Tracing, then Compression)
• Always call .with_state(state) last, after all routes and layers

Handlers

Handlers are async functions that receive extractors and return impl IntoResponse.

// src/handlers/users.rs
pub async fn register(
    State(state): State<AppState>,
    Json(dto): Json<CreateUserDto>,
) -> AppResult<(StatusCode, Json<AuthResponse>)> {
    let service = UserService::new(state.pool, state.config);
    let response = service.register(dto).await?;
    Ok((StatusCode::CREATED, Json(response)))
}

pub async fn list_users(
    State(state): State<AppState>,
    Query(pagination): Query<PaginationQuery>,
    _auth_user: AuthUser,
) -> AppResult<Json<Vec<UserResponse>>> {
    let service = UserService::new(state.pool, state.config);
    let users = service.list_users(pagination.page, pagination.per_page).await?;
    Ok(Json(users))
}

Handler Rules

• Keep handlers under 15 lines; delegate logic to services
• Always validate input via extractors or explicit .validate() calls
• Use AppResult<T> (a type alias for Result<T, AppError>) as the return type
• Extract AuthUser even if unused (_auth_user) to enforce authentication
• Return appropriate status codes: CREATED for POST, NO_CONTENT for DELETE

Extractors

Built-in Extractors

Extractor	Purpose	Example
State(state)	Shared application state	State(state): State<AppState>
Json(body)	JSON request body	Json(dto): Json<CreateUserDto>
Path(id)	URL path parameters	Path(id): Path<Uuid>
Query(params)	Query string parameters	Query(q): Query<PaginationQuery>
Extension(ext)	Request extensions	Extension(user): Extension<User>

Custom Extractor

// src/extractors/auth.rs
pub struct AuthUser {
    pub user_id: Uuid,
    pub email: String,
    pub role: String,
}

#[async_trait]
impl FromRequestParts<AppState> for AuthUser {
    type Rejection = AppError;

    async fn from_request_parts(
        parts: &mut Parts,
        state: &AppState,
    ) -> Result<Self, Self::Rejection> {
        let TypedHeader(Authorization(bearer)) = parts
            .extract::<TypedHeader<Authorization<Bearer>>>()
            .await
            .map_err(|_| AppError::Unauthorized("Missing authorization header".into()))?;

        let claims = verify_token(bearer.token(), &state.config.jwt_secret)?;

        Ok(AuthUser {
            user_id: claims.sub,
            email: claims.email,
            role: claims.role,
        })
    }
}

Extractor Rules

• Implement FromRequestParts (not FromRequest) when you only need headers/state
• Use FromRequest only when you need the body (consumes it)
• Extractor order in function signatures matters: body-consuming extractors must be last
• Always return your AppError type as the Rejection

State Management

#[derive(Clone)]
pub struct AppState {
    pub pool: sqlx::PgPool,
    pub config: Config,
}

impl AppState {
    pub fn new(pool: sqlx::PgPool, config: Config) -> Self {
        Self { pool, config }
    }
}

State Rules

• AppState must implement Clone (Axum requirement)
• Use sqlx::PgPool directly (it is Arc-wrapped internally)
• For mutable shared state, wrap in Arc<tokio::sync::RwLock<T>>
• Avoid std::sync::Mutex in async state; use tokio::sync::Mutex or RwLock
• Keep state lean: connection pools, config, caches -- not request-scoped data

Error Handling

// src/errors/app_error.rs
#[derive(Error, Debug)]
pub enum AppError {
    #[error("Not found: {0}")]
    NotFound(String),

    #[error("Bad request: {0}")]
    BadRequest(String),

    #[error("Unauthorized: {0}")]
    Unauthorized(String),

    #[error("Forbidden: {0}")]
    Forbidden(String),

    #[error("Conflict: {0}")]
    Conflict(String),

    #[error("Validation error: {0}")]
    Validation(String),

    #[error("Internal server error")]
    Internal(#[from] anyhow::Error),

    #[error("Database error")]
    Database(#[from] sqlx::Error),
}

impl IntoResponse for AppError {
    fn into_response(self) -> Response {
        let (status, message) = match &self {
            AppError::NotFound(msg) => (StatusCode::NOT_FOUND, msg.()),
            AppError::(msg) => (StatusCode::BAD_REQUEST, msg.()),
            AppError::(msg) => (StatusCode::UNAUTHORIZED, msg.()),
            AppError::(msg) => (StatusCode::FORBIDDEN, msg.()),
            AppError::(msg) => (StatusCode::CONFLICT, msg.()),
            AppError::(msg) => (StatusCode::UNPROCESSABLE_ENTITY, msg.()),
            AppError::(err) => {
                tracing::error!(, err);
                (StatusCode::INTERNAL_SERVER_ERROR, .())
            }
            AppError::(err) => {
                tracing::error!(, err);
                (StatusCode::INTERNAL_SERVER_ERROR, .())
            }
        };

          = (json!({
            : ,
            : { : status.(), : message }
        }));

        (status, body).()
    }
}

  <T> = <T, AppError>;

Error Handling Rules

• Always implement IntoResponse for your error type
• Never leak internal details (database errors, stack traces) to clients
• Log internal/database errors with tracing::error! before mapping to generic messages
• Use thiserror for enum definitions; anyhow for internal propagation
• Define AppResult<T> alias at the crate level for consistency

Middleware and Tower Integration

Route-Level Middleware

// Apply auth middleware to a route group
let protected = Router::new()
    .route("/api/v1/users", get(list_users))
    .layer(middleware::from_fn_with_state(state.clone(), auth_middleware));

Custom Middleware Function

pub async fn auth_middleware(
    State(state): State<AppState>,
    request: Request,
    next: Next,
) -> Result<Response, AppError> {
    let token = request.headers()
        .get(AUTHORIZATION)
        .and_then(|h| h.to_str().ok())
        .and_then(|h| h.strip_prefix("Bearer "))
        .ok_or_else(|| AppError::Unauthorized("Missing token".into()))?;

    verify_token(token, &state.config.jwt_secret)?;
    Ok(next.run(request).await)
}

Tower Layers (tower-http)

// Common layers -- add in create_router
.layer(TraceLayer::new_for_http())           // Request/response logging
.layer(CompressionLayer::new())              // Gzip response compression
.layer(CorsLayer::new().allow_origin(Any))   // CORS headers
.layer(TimeoutLayer::new(Duration::from_secs(30)))  // Request timeout

Middleware Rules

• Use middleware::from_fn_with_state for middleware needing AppState
• Use middleware::from_fn for stateless middleware
• Layer ordering: outermost layer executes first on request, last on response
• Always add TimeoutLayer to prevent slow requests from exhausting resources
• Use TraceLayer for observability in all environments

Commands

# Development
cargo run                           # Start server
cargo watch -x run                  # Watch mode (requires cargo-watch)
RUST_LOG=debug cargo run            # With debug logging

# Build
cargo build --release               # Production build
cargo check                         # Fast type-check

# Quality
cargo fmt                           # Format code
cargo clippy -- -D warnings         # Lint with deny
cargo audit                         # Vulnerability check

# Testing
cargo test                          # All tests
cargo test --test integration_tests # Integration only

# Database
sqlx migrate run                    # Run migrations
sqlx migrate add create_users       # Create new migration

Best Practices Summary

• Extractors: Use custom extractors for reusable validation; FromRequestParts for headers, FromRequest for body
• Error Handling: One AppError enum per service; implement IntoResponse; log internal errors, sanitize client messages
• Performance: Use connection pooling (SQLx), enable compression (tower-http), add timeouts, use async throughout
• Testing: Use axum-test for integration tests; mock repositories for unit tests; use a separate test database
• Security: Validate all inputs with validator; use parameterized queries (SQLx); never expose internal errors

Advanced Topics

For detailed handler examples, database integration, authentication flows, WebSocket support, and testing patterns, see:

• references/patterns.md -- Full CRUD handlers, repository pattern, JWT auth, WebSocket, integration testing

External References

• Axum Documentation
• Tower Documentation
• Tokio Documentation
• SQLx Documentation
• Axum Examples