Getting Started with Gemini

Google's Gemini models are natively multimodal, accepting text, images, audio, and video in a single request.

SDK Setup

import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI({ apiKey: process.env.GEMINI_API_KEY });

const response = await ai.models.generateContent({
  model: "gemini-2.5-flash",
  contents: "Explain how neural networks learn",
});
console.log(response.text);

Multimodal Inputs

Send images alongside text:

const response = await ai.models.generateContent({
  model: "gemini-2.5-flash",
  contents: [
    { text: "What's in this image?" },
    { inlineData: { mimeType: "image/jpeg", data: base64Image } }
  ],
});

Model Selection

Key Advantages

• Million-token context — process entire codebases or book-length documents
• Native multimodal — text, images, audio, video in one call
• Free tier — generous free usage for development and small projects
• Google ecosystem — integrates with Vertex AI, Cloud, and Search

Common Pitfall: Sending Base64 Images for Large Files

Inline base64 encoding works for images under ~1MB but becomes impractical for larger images or video files. For files over 1MB, use the File API to upload first:

const uploadedFile = await ai.files.upload({
  file: fs.createReadStream("large_image.jpg"),
  config: { mimeType: "image/jpeg" },
});

const response = await ai.models.generateContent({
  model: "gemini-2.5-flash",
  contents: [
    { text: "Describe this image in detail." },
    { fileData: { fileUri: uploadedFile.uri, mimeType: "image/jpeg" } },
  ],
});

Uploaded files persist for 48 hours and can be reused across multiple requests, avoiding redundant uploads.

Best Practice: Use System Instructions for Consistent Behavior

System instructions persist across the entire conversation and can significantly reduce prompt length for multi-turn interactions:

const response = await ai.models.generateContent({
  model: "gemini-2.5-flash",
  config: {
    systemInstruction: "You are a concise technical writer. Always respond in markdown. Keep answers under 300 words unless specifically asked for more detail.",
  },
  contents: "Explain REST vs GraphQL",
});

Named Product Reference: Vertex AI for Enterprise Gemini

For enterprise use, access Gemini through Google Cloud Vertex AI rather than the AI Studio API. Vertex provides VPC-SC support (keeps data inside your network perimeter), audit logging, custom IAM roles, and SLA-backed uptime guarantees. The SDK is nearly identical — replace GoogleGenAI with VertexAI from the @google-cloud/vertexai package.