@ragpipe/plugin-gemini

@ragpipe/plugin-gemini

Google Gemini embedding and generation plugin for ragpipe.

Install

pnpm add ragpipe @ragpipe/plugin-gemini

Usage

import { defineConfig } from "ragpipe";
import { geminiEmbedding, geminiGeneration } from "@ragpipe/plugin-gemini";

export default defineConfig({
  embedding: geminiEmbedding({
    apiKey: process.env.GEMINI_API_KEY ?? "",
    model: "gemini-embedding-001",
  }),
  generation: geminiGeneration({
    apiKey: process.env.GEMINI_API_KEY ?? "",
    model: "gemini-3.1-flash-lite-preview",
    systemPrompt: "Answer based on the provided context.",
  }),
  // ... vectorStore
});

API

geminiEmbedding(options)

Returns an EmbeddingPlugin that calls the Gemini Embedding API.

Option	Type	Default	Description
apiKey	string	—	Google AI API key (required)
model	string	-	Embedding model name (required)

• Dimensions: 3072
• Rate limit: 800ms between calls (built-in)
• Batch support: embedMany() uses batchEmbedContents for efficient bulk embedding

geminiGeneration(options)

Returns a GenerationPlugin that calls the Gemini Content Generation API.

Option	Type	Default	Description
apiKey	string	—	Google AI API key (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 Key

• Go to Google AI Studio
• Create an API key
• Set it as GEMINI_API_KEY in your environment

License

MIT