@ragpipe/plugin-cloudflare

@ragpipe/plugin-cloudflare

Cloudflare Workers AI embedding and generation plugin for ragpipe.

Install

pnpm add ragpipe @ragpipe/plugin-cloudflare

Usage

import { defineConfig } from "ragpipe";
import { cloudflareEmbedding, cloudflareGeneration } from "@ragpipe/plugin-cloudflare";

export default defineConfig({
  embedding: cloudflareEmbedding({
    accountId: process.env.CLOUDFLARE_ACCOUNT_ID!,
    apiToken: process.env.CLOUDFLARE_API_TOKEN!,
    model: "@cf/qwen/qwen3-embedding-0.6b",
  }),
  generation: cloudflareGeneration({
    accountId: process.env.CLOUDFLARE_ACCOUNT_ID!,
    apiToken: process.env.CLOUDFLARE_API_TOKEN!,
    model: "@cf/openai/gpt-oss-20b",
    systemPrompt: "Answer based on the provided context.",
  }),
  // ... vectorStore
});

API

cloudflareEmbedding(options)

Returns an EmbeddingPlugin that calls the Cloudflare Workers AI Embedding API.

Option	Type	Default	Description
accountId	string	—	Cloudflare account ID (required)
apiToken	string	—	Cloudflare API token (required)
model	string	—	Embedding model name (required)

• Batch support: embedMany() sends an array of texts in a single API call

cloudflareGeneration(options)

Returns a GenerationPlugin that calls the Cloudflare Workers AI Generation API.

Option	Type	Default	Description
accountId	string	—	Cloudflare account ID (required)
apiToken	string	—	Cloudflare API token (required)
model	string	—	Generation model name (required)
systemPrompt	string	"Answer based on the provided context."	Default system instruction

• Streaming: generateStream() returns an AsyncIterable<string> via SSE
• History: Pass { history } to include conversation context
• Per-call override: Pass { systemPrompt } at call time to override the default

Get an API Token

• Go to Cloudflare Dashboard
• Create a token with Workers AI read permission
• Set CLOUDFLARE_ACCOUNT_ID and CLOUDFLARE_API_TOKEN in your environment

License

MIT