Gemini
Installation
npm i llamaindex @llamaindex/google
pnpm add llamaindex @llamaindex/google
yarn add llamaindex @llamaindex/google
bun add llamaindex @llamaindex/google
Usage
import { gemini, GEMINI_MODEL } from "@llamaindex/google";
import { Settings } from "llamaindex";
Settings.llm = gemini({
model: GEMINI_MODEL.GEMINI_2_0_FLASH,
});
Usage with Vertex AI
To use Gemini via Vertex AI, you can specify the vertex configuration:
import { gemini, GEMINI_MODEL } from "@llamaindex/google";
const llm = gemini({
model: GEMINI_MODEL.GEMINI_2_0_FLASH,
vertex: {
project: "your-cloud-project", // required for Vertex AI
location: "us-central1", // required for Vertex AI
},
});
To authenticate for local development:
gcloud auth application-default login
To authenticate for production you'll have to use a service account. googleAuthOptions
has credentials
which might be useful for you.
Multimodal Usage
Gemini supports multimodal inputs including text, images, audio, and video:
import { gemini, GEMINI_MODEL } from "@llamaindex/google";
import fs from "fs";
const llm = gemini({ model: GEMINI_MODEL.GEMINI_2_0_FLASH });
const result = await llm.chat({
messages: [
{
role: "user",
content: [
{
type: "text",
text: "What's in this image?",
},
{
type: "image",
data: fs.readFileSync("./image.jpg").toString("base64"),
mimeType: "image/jpeg",
},
],
},
],
});
Tool Calling
Gemini supports function calling with tools:
import { gemini, GEMINI_MODEL } from "@llamaindex/google";
import { tool } from "llamaindex";
import { z } from "zod";
const llm = gemini({ model: GEMINI_MODEL.GEMINI_2_0_FLASH });
const result = await llm.chat({
messages: [
{
content: "What's the weather in Tokyo?",
role: "user",
},
],
tools: [
tool({
name: "weather",
description: "Get the weather",
parameters: z.object({
location: z.string().describe("The location to get the weather for"),
}),
execute: ({ location }) => {
return `The weather in ${location} is sunny and hot`;
},
}),
],
});
Live API (Real-time Conversations)
For real-time audio/video conversations using Gemini Live API.
The Live API is running directly in the frontend. That's why you have to generate an ephemeral key first on the server side and pass it to the frontend.
To use the Live API, make sure to pass apiVersion: "v1alpha"
to the httpOptions
.
import { gemini, GEMINI_MODEL } from "@llamaindex/google";
// Server-side: Generate ephemeral key
const serverLlm = gemini({
model: GEMINI_MODEL.GEMINI_2_0_FLASH_LIVE,
httpOptions: { apiVersion: "v1alpha" },
});
const ephemeralKey = await serverLlm.live.getEphemeralKey();
// Client-side: Use ephemeral key for Live API
const llm = gemini({
apiKey: ephemeralKey,
model: GEMINI_MODEL.GEMINI_2_0_FLASH_LIVE,
voiceName: "Zephyr",
httpOptions: { apiVersion: "v1alpha" },
});
const session = await llm.live.connect();
Load and index documents
For this example, we will use a single document. In a real-world scenario, you would have multiple documents to index.
import { Document, VectorStoreIndex } from "llamaindex";
const document = new Document({ text: essay, id_: "essay" });
const index = await VectorStoreIndex.fromDocuments([document]);
Query
const queryEngine = index.asQueryEngine();
const query = "What is the meaning of life?";
const results = await queryEngine.query({
query,
});
Full Example
import { gemini, GEMINI_MODEL } from "@llamaindex/google";
import { Document, VectorStoreIndex, Settings } from "llamaindex";
Settings.llm = gemini({
model: GEMINI_MODEL.GEMINI_2_0_FLASH,
});
async function main() {
const document = new Document({ text: essay, id_: "essay" });
// Load and index documents
const index = await VectorStoreIndex.fromDocuments([document]);
// Create a query engine
const queryEngine = index.asQueryEngine();
const query = "What is the meaning of life?";
// Query
const response = await queryEngine.query({
query,
});
// Log the response
console.log(response.response);
}
API Reference
Last updated on