Edge runtime

Cloudflare Workers, Vercel Edge, Deno Deploy — no OpenAI-SDK import, no `node:*` modules.

@cleverrouter/sdk/edge is a lighter entry that avoids the OpenAI SDK import. No node:* modules, no Buffer, no eval — everything works in runtimes that only expose Web APIs.

At a glance


Install	`bun add @cleverrouter/sdk@alpha`
Import path	`@cleverrouter/sdk/edge`
Runtimes	Cloudflare Workers, Vercel Edge, Deno Deploy
Available endpoints	`chat`, `stream`, `embed`, `rerank`, `transcribe`, `listModels`

Vercel Edge Function

app/api/chat/route.ts

import { createCleverRouter } from '@cleverrouter/sdk/edge';

export const runtime = 'edge';

export async function POST(req: Request) {
  const cr = createCleverRouter({
    apiKey: process.env.CLEVERROUTER_API_KEY!,
  });
  const { messages } = await req.json();

  const completion = await cr.chat({
    model: 'mistral/mistral-small-3.2',
    messages,
  });

  return Response.json(completion);
}

Streaming on the edge

Wrap the SDK's AsyncIterable into a ReadableStream and forward as SSE:

app/api/stream/route.ts

import { createCleverRouter } from '@cleverrouter/sdk/edge';

export const runtime = 'edge';

export async function POST(req: Request) {
  const cr = createCleverRouter({ apiKey: process.env.CLEVERROUTER_API_KEY! });
  const { messages } = await req.json();

  const stream = cr.stream({
    model: 'mistral/mistral-small-3.2',
    messages,
  });

  const encoder = new TextEncoder();
  const body = new ReadableStream({
    async start(controller) {
      try {
        for await (const chunk of stream) {
          const text = chunk.choices[0]?.delta?.content ?? '';
          if (text) {
            controller.enqueue(
              encoder.encode(`data: ${JSON.stringify({ delta: text })}\n\n`),
            );
          }
        }
        controller.enqueue(encoder.encode('data: [DONE]\n\n'));
      } catch (err) {
        controller.error(err);
      } finally {
        controller.close();
      }
    },
  });

  return new Response(body, {
    headers: {
      'Content-Type': 'text/event-stream',
      'Cache-Control': 'no-cache, no-transform',
      'X-Accel-Buffering': 'no',
    },
  });
}

Cloudflare Workers

worker.ts

import { createCleverRouter } from '@cleverrouter/sdk/edge';

export default {
  async fetch(req: Request, env: { CLEVERROUTER_API_KEY: string }) {
    const cr = createCleverRouter({ apiKey: env.CLEVERROUTER_API_KEY });
    const { messages } = await req.json();
    return Response.json(
      await cr.chat({ model: 'mistral/mistral-small-3.2', messages }),
    );
  },
};

The package's exports map advertises ./edge with the worker condition, so wrangler picks the right artifact automatically.

Deno Deploy

main.ts

import { createCleverRouter } from 'npm:@cleverrouter/sdk@alpha/edge';

const cr = createCleverRouter({ apiKey: Deno.env.get('CLEVERROUTER_API_KEY')! });

Deno.serve(async (req) => {
  const { messages } = await req.json();
  return Response.json(
    await cr.chat({ model: 'mistral/mistral-small-3.2', messages }),
  );
});

What's missing vs the main entry

Feature	`@cleverrouter/sdk`	`@cleverrouter/sdk/edge`
`cr.openai()` shim	yes	no
`chat`, `stream`	yes	yes
`embed`, `rerank`	yes	yes
`transcribe`	yes	yes
`listModels`	yes	yes
React hooks	via `/react`	n/a
Bundle size impact	+ `openai` peer dep	minimal

Mix and match per file

Same SDK package — import @cleverrouter/sdk in Node.js / Bun routes and @cleverrouter/sdk/edge from edge routes. Types and error classes are shared.

Common gotchas

process.env doesn't exist in Cloudflare Workers. Use the env argument that fetch receives.
Buffer doesn't exist on the edge. Use Uint8Array and TextEncoder/TextDecoder — the SDK already does.
Edge fetch quotas. Vercel and Cloudflare cap streamed connections per worker. Long-lived chat streams should run on Node.js or Bun routes — the edge is great for one-shot completions and proxy patterns.

SDK overview
Streaming — wire format details.
React hooks — pair with /api/chat on the edge.

On this page