@anvia/gemini: Getting Started

Install

pnpm add @anvia/gemini @anvia/core

Configure credentials

Set GEMINI_API_KEY in the server environment. Keep provider keys on the server side; browser clients should call an application route that owns the model request. For Vertex AI, construct GeminiClient with vertexai: true, project, and location instead of an API key.

Minimum setup

import { AgentBuilder } from "@anvia/core";
import { GeminiClient } from "@anvia/gemini";

const client = new GeminiClient({
  apiKey: process.env.GEMINI_API_KEY,
});

const model = client.completionModel("gemini-2.5-flash");

const agent = new AgentBuilder("assistant", model)
  .instructions("Answer clearly and concisely.")
  .build();

const response = await agent.prompt("Summarize this ticket.").send();
console.log(response.output);

Other model factories

@anvia/gemini also exposes an embedding model factory:

const embeddings = client.embeddingModel("gemini-embedding-001");
const vectors = await embeddings.embedTexts(["Refunds take five business days."]);

Next step

Continue with Usage Patterns.