Zig package to enable a zig-based 'function as a service' or 'Lambda function'
Go to file
Emil Lerch 5f1b1a52be
AWS-Zig Build / build-zig-0.11.0-amd64-host (push) Successful in 1m6s Details
add resource usage stats when built in debug mode
2024-02-25 17:04:08 -08:00
.gitea/workflows try wasmtime again 2024-01-09 14:34:49 -08:00
src add resource usage stats when built in debug mode 2024-02-25 17:04:08 -08:00
.gitignore it is alive 2023-09-16 08:14:11 -07:00
LICENSE it is alive 2023-09-16 08:14:11 -07:00
README.md update build options module name and readme 2023-10-28 15:51:41 -07:00
build.zig re-enable lib tests 2024-01-09 14:42:18 -08:00
build.zig.zon allow handler more control in status reporting 2023-10-23 12:49:18 -07:00


"Univeral Lambda" for Zig

This a Zig 0.11 project intended to be used as a package to turn a zig program into a function that can be run as:

  • A command line executable
  • A standalone web server
  • An AWS Lambda function
  • A shared library in flexilib
  • Cloudflare
  • etc

Usage - Development

From an empty directory, with Zig 0.11 installed:

zig init-exe

Create a build.zig.zon with the following contents:

    .name = "univeral-zig-example",
    .version = "0.0.1",

    .dependencies = .{
        .universal_lambda_build = .{
            .url = "https://git.lerch.org/lobo/universal-lambda-zig/archive/07366606696081f324591b66ab7a9a176a38424c.tar.gz",
            .hash = "122049daa19f61d778a79ffb82c64775ca5132ee5c4797d7f7d76667ab82593917cd",
        .flexilib = .{
            .url = "https://git.lerch.org/lobo/flexilib/archive/c44ad2ba84df735421bef23a2ad612968fb50f06.tar.gz",
            .hash = "122051fdfeefdd75653d3dd678c8aa297150c2893f5fad0728e0d953481383690dbc",

Due to limitations in the build apis related to relative file paths, the dependency name currently must be "universal_lambda_build". Also, note that the flexilib dependency is required at all times. This requirement may go away with zig 0.12 (see #17135) and/or changes to this library.


  • Add an import at the top:
const universal_lambda = @import("universal_lambda_build");
  • Set the return of the build function to return !void rather than void
  • Add a line to the build script, after any modules are used, but otherwise just after adding the exe is fine:
try universal_lambda.configureBuild(b, exe);

This will provide most of the magic functionality of the package, including several new build steps to manage the system, and a new import to be used. For testing, it is also advisable to add the modules to your tests by adding a line like so:

_ = try universal_lambda.addModules(b, main_tests);


The build changes above will add several modules:

  • universal_lambda_handler: Main import, used to register your handler
  • universal_lambda_interface: Contains the context type used in the handler function
  • flexilib-interface: Used as a dependency of the handler. Not normally needed

Add imports for the handler registration and interface:

const universal_lambda = @import("universal_lambda_handler");
const universal_lambda_interface = @import("universal_lambda_interface");

Another module universal_lambda_build_options is available if access to the environment is needed. The build type is stored under a build_type variable.

Add a handler to be executed. The handler must follow this signature:

pub fn handler(allocator: std.mem.Allocator, event_data: []const u8, context: universal_lambda_interface.Context) ![]const u8

Your main function should return !u8. Let the package know about your handler in your main function, like so:

return try universal_lambda.run(null, handler);

The first parameter above is an allocator. If you have a specific handler you would like to use, you may specify it. Otherwise, an appropriate allocator will be created and used. Currently this is an ArenaAllocator wrapped around an appropriate base allocator, so your handler does not require deallocation.

A fully working example of usage is at https://git.lerch.org/lobo/universal-lambda-example/.

Usage - Building

The build configuration will add the following build steps when building with Linux:

  awslambda_package            Package the function
  awslambda_deploy             Deploy the function
  awslambda_iam                Create/Get IAM role for function
  awslambda_run                Run the app in AWS lambda
  cloudflare                   Deploy as Cloudflare worker (must be compiled with -Dtarget=wasm32-wasi)
  flexilib                     Create a flexilib dynamic library
  standalone_server            Run the function in its own web server

AWS Lambda is not currently available if building with other operating systems, as that set of build steps utilize system commands using the AWS CLI. This is likely to change in the future to enable other operating systems. All other build steps are available for all targets.

Note that AWS Lambda will require that credentials are established using the same methods as checked by the AWS CLI and the AWS CLI is installed.

If using Cloudflare deployment, either CLOUDFLARE_API_TOKEN or CLOUDFLARE_EMAIL/CLOUDFLARE_API_KEY environment variables must be set for successful deployment.

To run as an executable, a simple zig build will build, or zig build run will run as expected. zig build standalone_server run will also build/run as a standalone web server.


Limitations include standalone web server port customization and linux/aws cli requirements for Linux.

Also, within the context, AWS Lambda is unable to provide proper method, target, and headers for the request. This may be important for routing purposes. Suggestion here is to use API Gateway and pass these parameters through the event_data content.

Lastly, support for specifying multiple targets in the downstream (your) application is somewhat spotty. For example, zig build standalone_server run works fine. However, zig build test flexilib is broken.