// the find
anthdm/raptor
Create, Deploy, and Run your applications on the edge
Raptor is a Go framework for deploying WASM functions to an edge runtime, using an actor model (looks like proto.actor under internal/actrs) for isolation and supervision between function instances. It's aimed at people who want a self-hosted FaaS layer that runs arbitrary WASM (Go or JS via an embedded SpiderMonkey build) instead of renting Lambda or Cloudflare Workers.
Running untrusted code as WASM gives you real sandboxing without containers, and the actor model is a sensible fit for supervising many short-lived function instances with restart/fault isolation built in rather than bolted on. Supporting both Go-compiled WASM and JS (via a vendored SpiderMonkey wasm blob) means it's not locked to one language. The codebase is split cleanly into ingress/runtime/api/storage packages, which is a reasonable shape for eventually running these as separate deployable services.
Last commit is January 2024 and the README states outright that it's 'work in progress and rough on the edges' with no install instructions written yet — this is an abandoned prototype, not something to build on. The deploy endpoint (POST /endpoint/<id>/deploy) takes a raw WASM blob with no auth or signing mentioned anywhere in the docs, which is a real problem for anything described as 'deploy to the edge.' Test coverage is thin — only a handful of _test.go files across the whole tree, nothing for the actor or wasmserver logic that actually executes untrusted code. Checking a compiled js.wasm binary straight into git (twice, under cmd/cli and internal/spidermonkey) instead of building it in CI is a maintenance smell and bloats the repo.