Legacy community-maintained package to support using Rust inside Vercel Functions as a Runtime.
Important
This community runtime is now deprecated in favor of the official Vercel Rust runtime implementation.
For new issues please open them in the Vercel CLI repository.
The new runtime crate is released with 2.x onwards. This repository's legacy builder is intended for use with the legacy runtime 1.x.
Templates
Support Status Deployments using this legacy builder and runtime will continue to function, but will not receive support or updates anymore. We recommend migrating to the new official runtime for continued support and new features.
Legacy Documentation
Please ensure Vercel CLI and the Rust toolchain is already installed on your system. We recommended setting up Rust with rustup.
Step 1 β Add a vercel.json file to your project.
{
"functions": {
"api/**/*.rs": {
"runtime": "vercel-rust@4.0.9"
}
}
}This turns every file matching api/**/*.rs into a Vercel Function.
Note: The npm dependency
vercel-rustdefined in functionsdoes not have to be installed manually.
Step 2 β Create a function. As an example, here is api/handler.rs.
use serde_json::json;use vercel_runtime::{run,Body,Error,Request,Response,StatusCode};#[tokio::main]asyncfnmain() -> Result<(),Error>{run(handler).await}pubasyncfnhandler(_req:Request) -> Result<Response<Body>,Error>{Ok(Response::builder().status(StatusCode::OK).header("Content-Type","application/json").body(json!({"message":"δ½ ε₯½οΌδΈη"}).to_string().into(),)?)}Step 3 β Create a Cargo.toml in the root directory of your project.
[package]
name = "my-vercel-api"version = "0.1.0"edition = "2021"
[dependencies]
tokio = { version = "1", features = ["macros"] }
serde_json = { version = "1", features = ["raw_value"] }
# Documentation: https://docs.rs/vercel_runtime/latest/vercel_runtimevercel_runtime = { version = "1" }
# You can specify a library for shared logic here (optional)# [lib]# path = "src-rs/lib.rs"# Each handler has to be specified as [[bin]]
[[bin]]
name = "handler"path = "api/handler.rs"# Note that you need to provide unique names for each binary:# [[bin]]# name = "user-id"# path = "api/user/[id].rs"## [[bin]]# name = "group-id"# path = "api/group/[id].rs"Step 4 β Create a .vercelignore in the root directory of your project to ignore build artifacts.
target/Step 5 β You're all set. Run vercel dev to develop your project locally. You can connect a Git repository to Vercel, or use vercel to start deploying your project on Vercel.
An example on how this can be achieved is using a rust-toolchain file adjacent to your Cargo.toml. Please refer to Rust Documentation for more details.
By default builder module supports installing dependencies defined in the Cargo.toml file.
More system dependencies can be installed at build time with the presence of a shell build.sh file in the root directory of your project.
When creating a prebuilt deployment, the build output must be for x86_64 linux. To do this, create a Cargo build configuration at .cargo/config.toml with the following contents:
[build]
target = "x86_64-unknown-linux-musl"# Uncomment below to support Rust cross-compilation from macOS to Linux# Follow these installation instructions: https://github.com/chinedufn/cross-compile-rust-from-mac-to-linux# [target.x86_64-unknown-linux-musl]# linker = "x86_64-unknown-linux-gnu-gcc"You then can build the file and trigger the deployment via Vercel CLI.
vercel build && vercel deploy --prebuiltUnfortunately, the AWS Lambda Runtime for Rust relies (tangentially) on proc_macro, which won't compile on musl targets. Without musl, all linking must be dynamic. If you have a crate that relies on system libraries like postgres or mysql, you can include those library files with the includeFiles config option and set the proper environment variables, config, etc. that you need to get the library to compile.
For more information, please see this issue.
This feature allows you to bundle all of your routes into a single deployed Vercel function.
This serves to optimize cold starts, as lambda functions are reused as much as possible.
In addition, this has the benefit of only needing to annotate a single [[bin]] in your Cargo.toml.
To enable this behaviour, take the following steps:
Step 1 β Create a api/main.rs.
use vercel_runtime::{bundled_api, run,Body,Error,Request,Response};#[tokio::main]asyncfnmain() -> Result<(),Error>{run(handler).await}// The proc macro `bundled_api` injects a router for all `api/**/*.rs` handler files .// If you are using cargo workspaces (like `examples/route-merge` in this repository),// then an additional `path` argument must be passed to the macro. E.g.// #[bundled_api( path = "examples/route-merge" )]#[bundled_api]pubasyncfnhandler(req:Request) -> Result<Response<Body>,Error>{}Step 2 β Change your vercel.json to only specify your api/main.rs file.
{
"functions": {
"api/main.rs": {
"runtime": "vercel-rust@4.0.9"
}
}
}Step 3 β Change your Cargo.toml to specify the binary for main.rs.
[[bin]]
name = "main"path = "api/main.rs"Step 4 β Add a handler function to each route in api/**.
// Example api/foo.rsuse vercel_runtime::{Body,Error,Request,Response,StatusCode};pubasyncfnhandler(_req:Request) -> Result<Response<Body>,Error>{Ok(Response::builder().status(StatusCode::OK).header("Content-Type","application/json").body(Body::Text("Route is /api/foo".into()))?)}Since this project contains both Rust and Node.js code, you need to install the relevant dependencies. If you're only working on the TypeScript side, you only need to install those dependencies (and vice-versa).
# install node dependencies
pnpm install
# install cargo dependencies
cargo fetchThe npm module vercel-rust is implementing an interface which is primarily taking care of spawning a development server, caching between consecutive builds, and running the compilation. You can read more about the in-depths of implementing a builder here.
The crate vercel_runtime is what you will consume in your Rust functions. As the name suggests, the runtime crate takes care of everything that happens during run-time. In specific it takes care of creating a Tower service, which expects a specific handler signature. The flow of an invocation can be visualized as the following:
graph TD
A["Function Invocation"] --> |"process_request(event: InvocationEvent<VercelEvent>) β Request"| B[Request]
B --> |"handler_fn(req: Request) β Future<Output = Result<Response<Body>, Error>>"| C["Runtime calls handler_fn"]
C --> |"Ok(r) => process_response(r)"| D["Response"]
The crate vercel_runtime_macro enables supporting our API bundling feature. This procedural macro matches all api/**/*.rs routes, imports their respective handlers, and injects router logic to call the correct handlers during runtime.
The crate vercel_runtime_router contains routing logic that is injected for our API bundling feature.
If you are looking for the legacy runtime instructions using vercel_lambda see tree/a9495a0.