capy-framework 0.10.0

An opinionated full-stack Rust web framework with strong conventions and sensible defaults.
Documentation

Capy

A calmer way to build web apps in Rust.

Strong conventions. Sensible defaults. No framework assembly required.

Capy is an opinionated full-stack Rust web framework with strong conventions and sensible defaults. It combines a Rust backend with a first-class frontend workflow so you can start building the application instead of assembling the framework.

[dependencies]
capy = { package = "capy-framework", version = "0.10.0" }

Application code normally starts with use capy::prelude::*;. Native Axum, SeaORM, Tokio and Serde APIs remain available through named re-exports. Generate a complete application with capy new blog using capy-cli.

Generated resources keep entities separate from serialized DTOs. Use bounded Page<ResponseDto> collections and explicit public DTOs; see the executable tutorial below. Production uses JSON request logs with request ID, method, route, status, and latency. Set RUST_LOG to override the default INFO filter. Responses include x-request-id; internal errors are sanitized. Prefer running the compiled release binary in production.

MIT licensed. See LICENSE. This is an initial developer preview, not a production-readiness claim.

New applications use App::api_routes(routes) to mount paths under /api and App::frontend("frontend/dist") for static assets and SPA fallback. App::routes continues to register unprefixed routes. Explicit web routes and API handlers take precedence; API 404s never return the SPA. Missing static assets return 404, while GET/HEAD browser paths without a file extension fall back to index.html.

capy dev runs the backend and Vite together; capy build produces frontend assets and a release binary. The production backend requires no Node server. Deploy the frontend/dist directory with the binary and run from the project root with APP_ENVIRONMENT=production; missing frontend builds fail before database startup. into_router constructs only the HTTP service; run/run_command perform startup validation. Existing backend-only apps keep their current behavior.

Optional identity

Capy 0.6.0 introduced experimental cookie sessions through the auth feature. Run capy make:auth to install conventional SeaORM migrations, safe DTOs, CSRF-aware React helpers and an application-owned reset-delivery hook. Use AuthUser or OptionalAuth and ordinary Policy implementations for protected resources. Production requires HTTPS and a real reset-delivery adapter; generated local delivery is development-only. See the auth API.

Build a Complete Feature

Run capy make:auth --frontend then capy make:resource Post title:string slug:string:unique body:text published:boolean=false --frontend. Apply capy db:migrate, capy db:seed, then capy dev. The primary tutorial connects generated CRUD, auth pages, typed client calls, TestApp/TestSession and fixtures. The lower-level API tutorial follows below.

Your first Posts API

This tutorial targets Capy 0.10.0, with Rust 1.88 or later. Keep the CLI and framework on the same release. For the primary generated full-stack workflow, start with Build a Complete Feature.

It creates a JSON API with SQLite first, then shows PostgreSQL and generated React contracts. All application imports come from capy; no direct Axum, SeaORM, Serde, Tokio or Schemars dependencies are needed.

1. Install and create

Install Rust through rustup, with a native linker/toolchain for your OS. Install the matching CLI and create an application:

rustup component add rustfmt
cargo install capy-cli --version 0.10.0 --locked
capy --version
capy new blog --api
cd blog

The generated Cargo.toml uses capy-framework 0.10.0 from crates.io, including its testing feature in development dependencies. No local source override is needed.

Cargo installs the executable in $HOME/.cargo/bin; make sure that directory is on PATH (rustup normally configures it). The first installation/application build compiles dependencies and takes longer than later runs. A minimal Linux image needs a C compiler/linker, such as Debian/Ubuntu's build-essential package. SQLite is bundled; no SQLite service or PostgreSQL client library is required.

2. Choose a database

The generated SQLite configuration works immediately:

DATABASE_URL=sqlite://database/app.sqlite?mode=rwc

You do not need to create a .env file for this default. For PostgreSQL, create a disposable tutorial database. One option is Docker:

docker run --rm -d --name capy-postgres \
  -e POSTGRES_PASSWORD=capy-test -e POSTGRES_DB=capy_blog \
  -p 127.0.0.1:5432:5432 postgres:17-alpine
export DATABASE_URL=postgres://postgres:capy-test@127.0.0.1:5432/capy_blog

Wait for docker exec capy-postgres pg_isready -U postgres -d capy_blog to report that connections are accepted. Keep DATABASE_URL in the shell used for migration and serving, or put it in an ignored .env file. The same code and migrations below work on both databases. Use application-specific credentials outside this tutorial.

3. Generate the components

capy make:model Post
capy make:migration create_posts_table
capy make:request CreatePostRequest
capy make:controller PostController

These commands register compilable skeletons. Add your domain fields and handlers by replacing the indicated files with the complete examples below. Keep the registrations generated in each mod.rs file.

4. Add the model field

Replace app/models/post.rs. The generator already supplies serialization; we add one title column. The Post entity and PostModel aliases are already registered.

use capy::sea_orm::{self, entity::prelude::*};

#[derive(Clone, Debug, PartialEq, Eq, DeriveEntityModel)]
#[sea_orm(table_name = "posts")]
pub struct Model {
    #[sea_orm(primary_key)]
    pub id: i32,
    pub title: String,
}

#[derive(Copy, Clone, Debug, EnumIter, DeriveRelation)]
pub enum Relation {}

impl ActiveModelBehavior for ActiveModel {}

5. Define the migration

Replace the generated file ending in _create_posts_table.rs under database/migrations/. Keep its generated filename and the migrator registry. Apply migrations explicitly; starting the server never changes the schema.

use capy::migration::{self as sea_orm_migration, prelude::*};

#[derive(DeriveMigrationName)]
pub struct Migration;

#[async_trait::async_trait]
impl MigrationTrait for Migration {
    async fn up(&self, manager: &SchemaManager) -> Result<(), DbErr> {
        manager
            .create_table(
                Table::create()
                    .table(Alias::new("posts"))
                    .col(
                        ColumnDef::new(Alias::new("id"))
                            .integer()
                            .not_null()
                            .auto_increment()
                            .primary_key(),
                    )
                    .col(
                        ColumnDef::new(Alias::new("title"))
                            .string_len(200)
                            .not_null(),
                    )
                    .to_owned(),
            )
            .await
    }

    async fn down(&self, manager: &SchemaManager) -> Result<(), DbErr> {
        manager
            .drop_table(Table::drop().table(Alias::new("posts")).to_owned())
            .await
    }
}

6. Validate input

Replace app/requests/create_post_request.rs. Both create and update require a title between 2 and 200 characters. Invalid input receives HTTP 422 before the handler runs.

use capy::prelude::*;

#[derive(Debug, Deserialize, Validate, ApiSchema)]
#[serde(crate = "capy::serde")]
#[schemars(crate = "capy::schema")]
#[validate(crate = "capy::validator")]
pub struct CreatePostRequest {
    #[validate(length(min = 2, max = 200))]
    pub title: String,
}

7. Implement the controller

Replace app/controllers/post_controller.rs. Context supplies the typed database connection. Native SeaORM errors convert through ?; missing rows become JSON 404s. Query<Pagination> defaults to page 1 with 20 items and rejects page sizes outside 1–100. Try /api/posts?page=2&per_page=20; lists return data plus meta with page, per_page, total, and last_page. Empty collections have last_page: 1. The explicit PostResponse keeps future private database columns out of the API.

use crate::app::{
    models::{post, Post, PostModel},
    requests::create_post_request::CreatePostRequest,
};
use capy::{
    prelude::*,
    sea_orm::{ActiveModelTrait, EntityTrait, IntoActiveModel, PaginatorTrait, QueryOrder, Set},
};

pub struct PostController;

#[derive(Serialize, ApiSchema)]
#[serde(crate = "capy::serde")]
#[schemars(crate = "capy::schema")]
pub struct PostResponse {
    pub id: i32,
    pub title: String,
}

impl From<PostModel> for PostResponse {
    fn from(post: PostModel) -> Self {
        Self {
            id: post.id,
            title: post.title,
        }
    }
}

impl PostController {
    pub async fn index(
        Context(ctx): Context,
        Query(page): Query<Pagination>,
    ) -> Result<Json<Page<PostResponse>>> {
        let query = Post::find()
            .order_by_asc(post::Column::Id)
            .paginate(&ctx.db, page.per_page());
        let total = query.num_items().await?;
        let posts = query.fetch_page(page.page() - 1).await?;
        Ok(Json(Page::new(
            posts.into_iter().map(PostResponse::from).collect(),
            page,
            total,
        )))
    }

    pub async fn store(
        Context(ctx): Context,
        Validated(request): Validated<CreatePostRequest>,
    ) -> Result<Created<PostResponse>> {
        Ok(Created(
            post::ActiveModel {
                title: Set(request.title),
                ..Default::default()
            }
            .insert(&ctx.db)
            .await?
            .into(),
        ))
    }

    pub async fn show(Context(ctx): Context, Path(id): Path<i32>) -> Result<Json<PostResponse>> {
        Ok(Json(
            Post::find_by_id(id)
                .one(&ctx.db)
                .await?
                .ok_or(AppError::NotFound)?
                .into(),
        ))
    }

    pub async fn update(
        Context(ctx): Context,
        Path(id): Path<i32>,
        Validated(request): Validated<CreatePostRequest>,
    ) -> Result<Json<PostResponse>> {
        let mut post = Post::find_by_id(id)
            .one(&ctx.db)
            .await?
            .ok_or(AppError::NotFound)?
            .into_active_model();
        post.title = Set(request.title);
        Ok(Json(post.update(&ctx.db).await?.into()))
    }

    pub async fn destroy(Context(ctx): Context, Path(id): Path<i32>) -> Result<NoContent> {
        if Post::delete_by_id(id).exec(&ctx.db).await?.rows_affected == 0 {
            return Err(AppError::NotFound);
        }
        Ok(NoContent)
    }
}

8. Register the routes

Replace routes/api.rs. Generator module registration does not implicitly add HTTP routes. The generated application mounts these paths under /api via App::api_routes.

use crate::app::{
    controllers::post_controller::{PostController, PostResponse},
    requests::create_post_request::CreatePostRequest,
};
use capy::prelude::*;

pub fn routes() -> Routes {
    Routes::new()
        .get("/health", capy::health)
        .operation("health")
        .response::<capy::HealthResponse>()
        .get("/ready", capy::ready)
        .operation("readiness")
        .response::<capy::ReadinessResponse>()
        .error_response::<capy::ReadinessResponse>(503)
        .get("/posts", PostController::index)
        .operation("listPosts")
        .query::<Pagination>()
        .response::<Paginated<PostResponse>>()
        .tag("Posts")
        .error(500)
        .post("/posts", PostController::store)
        .operation("createPost")
        .body::<CreatePostRequest>()
        .created::<PostResponse>()
        .tag("Posts")
        .error(500)
        .get("/posts/{id}", PostController::show)
        .operation("showPost")
        .path::<i32>("id")
        .response::<PostResponse>()
        .error(404)
        .put("/posts/{id}", PostController::update)
        .operation("updatePost")
        .path::<i32>("id")
        .body::<CreatePostRequest>()
        .response::<PostResponse>()
        .error(404)
        .delete("/posts/{id}", PostController::destroy)
        .operation("deletePost")
        .path::<i32>("id")
        .no_content()
        .error(404)
}

9. Migrate and run

cargo fmt --check
cargo check --all-targets
capy migrate
capy migrate:status
capy routes --api
capy openapi
capy serve

The server listens at http://127.0.0.1:8000. GET /api/health returns HTTP 200 and {"status":"ok"}. Use APP_PORT=8001 capy serve if port 8000 is occupied.

10. Call the API

In another terminal, while capy serve is running:

curl -i http://127.0.0.1:8000/api/health
curl -i -X POST http://127.0.0.1:8000/api/posts \
  -H 'Content-Type: application/json' -d '{"title":"Hello, Capy!"}'
curl -i http://127.0.0.1:8000/api/posts
curl -i http://127.0.0.1:8000/api/posts/1
curl -i -X PUT http://127.0.0.1:8000/api/posts/1 \
  -H 'Content-Type: application/json' -d '{"title":"Updated post"}'
curl -i -X POST http://127.0.0.1:8000/api/posts \
  -H 'Content-Type: application/json' -d '{"title":"x"}'
curl -i -X DELETE http://127.0.0.1:8000/api/posts/1

On a fresh database, the created ID is 1; otherwise use the ID returned by POST. Expect 201 on create, 200 on list/show/update, 422 for the short title, and 204 on delete. Fetching a deleted ID gives {"message":"Resource not found"} with 404. Every response has x-request-id; supply your own valid ID to correlate logs. Internal database failures receive a sanitized JSON 500.

Stop the server with Ctrl-C. Unix SIGTERM to the capy serve PID also shuts down and reaps its server. In production, run the compiled application binary directly.

11. Test the application

Create tests/posts.rs. These tests use an isolated in-memory SQLite database, regardless of the server database chosen above.

use blog::application;
use capy::{prelude::*, testing::TestApp};

#[capy::tokio::test(crate = "capy::tokio")]
async fn posts_are_created_and_validated() -> Result<()> {
    let app = TestApp::boot(application()).await?;
    let created = app
        .post("/api/posts")
        .json(json!({"title": "Hello, Capy!"}))
        .send()
        .await;
    created.assert_created();
    let post: Value = created.json();
    assert_eq!(post["title"], "Hello, Capy!");
    assert!(created.headers.contains_key("x-request-id"));
    app.get("/api/posts").send().await.assert_ok().assert_json(
        json!({"data":[post],"meta":{"page":1,"per_page":20,"total":1,"last_page":1}}),
    );
    app.post("/api/posts")
        .json(json!({"title": "x"}))
        .send()
        .await
        .assert_status(capy::axum::http::StatusCode::UNPROCESSABLE_ENTITY);
    Ok(())
}
cargo fmt --check
cargo clippy --all-targets -- -D warnings
cargo test

You can also use capy test. To practice migration rollback after stopping the server, run capy migrate:rollback, check capy migrate:status, then capy migrate to reapply. Rolling back this migration deletes its table and data.

If you started the disposable PostgreSQL container, remove it when finished:

docker stop capy-postgres

See the API documentation for configuration types and native escape hatches. This release is an initial developer preview.

12. Use the contract from React

For a full-stack project, start with capy new blog (omit --api) and follow the same Rust steps. capy types writes generated/openapi.json and frontend/src/generated/api-types.ts. Run pnpm install --frozen-lockfile inside frontend/ first, or let capy dev install dependencies and generate the contracts after the backend compiles.

import { api } from '@/lib/api'
import type { CreatePostRequest } from '@/generated/api-types'

const payload: CreatePostRequest = { title: 'Hello, Capy!' }
const post = await api.post('/api/posts', {
  body: payload,
})
const page = await api.get('/api/posts', { params: { query: { page: 1 } } })
await api.delete('/api/posts/{id}', { params: { path: { id: post.id } } })

Keep generated files in source control. CI runs capy types --check after installing frontend dependencies, followed by the Rust tests and pnpm build. A stale contract causes a non-zero exit without rewriting files. capy dev regenerates after successful Rust builds, and capy build generates before compiling React.

These are compile-time JSON contracts, not runtime response validators. Database models stay separate from response DTOs. See API contracts for Serde semantics, validation metadata, API documentation, and custom wire types.