URL: /geist/guides/edge-functions

---
title: Edge functions
description: Run TypeScript, Go, or Python at 280+ POPs with sub-30ms cold starts.
icon: zap
---

Edge functions run as close to your users as the network allows. Halo isolates each invocation in a V8 isolate (TS) or a microVM (Go/Python), and routes requests to the nearest healthy POP.

## Create a function

```ts api/hello.ts
export default function handler(req: Request) {
  return new Response("hi from " + req.headers.get("x-halo-region"));
}
```

The file path becomes the route. `api/hello.ts` is reachable at `/api/hello` after deploy.

## Per-region pinning

By default, functions run in every POP. Pin to a region:

```ts
export const config = {
  regions: ["iad1", "fra1"],
};
```

<Tabs>
  <Tab title="TypeScript">
    Runtime: V8 isolates. Cold start: ~5–15ms. Memory: 128 MB max.
  </Tab>
  <Tab title="Go">
    Runtime: Firecracker microVM. Cold start: ~25–50ms. Memory: up to 1 GB.
  </Tab>
  <Tab title="Python">
    Runtime: Firecracker microVM. Cold start: ~30–60ms. Memory: up to 1 GB.
  </Tab>
</Tabs>

## Streaming responses

```ts
export default async function handler() {
  const stream = new ReadableStream({
    start(controller) {
      controller.enqueue(new TextEncoder().encode("first chunk\n"));
      setTimeout(() => {
        controller.enqueue(new TextEncoder().encode("second chunk\n"));
        controller.close();
      }, 100);
    },
  });
  return new Response(stream);
}
```

Halo flushes each chunk to the client as soon as it's enqueued.

<Note>
  Edge functions can't bind raw TCP sockets. For databases, use HTTP-native drivers (Neon, PlanetScale serverless, Upstash) or hit them through a regional service.
</Note>