DCXTribeDCXTribe
Back to library
AI
tutorial

Slash Multi-Query Costs Using Gemini Context Caching

Drastically reduce API costs and latency for repetitive, long-form queries.

Use Gemini's Context Caching to store massive reference documents on Google's servers once, rather than re-sending them with every API call.

Google Gemini

The Scenario

You are building a tool for users to ask questions about a specific 50-page government proposal or legal contract. You want to avoid the high cost and latency of sending the same document text with every user message.

Before & after

The old way

Previously, you had to re-upload and re-process the entire 50-page PDF with every single follow-up question. (Rough time cost: 30-60 seconds of latency + high repeated token costs).

With AI

By using Context Caching, you only pay to store the documents once; Subsequent queries process instantly and cost significantly less. (Rough time cost: 30 seconds to set up the cache).

The Prompt

// Using the Gemini Node.js SDK
import { GoogleAICacheManager } from "@google/generative-ai/server";

const cacheManager = new GoogleAICacheManager(process.env.API_KEY);
const cache = await cacheManager.create({
  model: "models/gemini-1.5-pro-001",
  contents: [
    { role: "user", parts: [{ text: "[INSERT_LONG_DOCUMENT_TEXT_HERE]" }] }
  ],
  ttlSeconds: 3600, // Valid for 1 hour
});

console.log(`Cache created: ${cache.name}`);

Context Caching is particularly effective for repeated queries against massive datasets like legal documents, technical manuals, or political manifestos where the core source doesn't change.

Source

Release notes  |  Gemini API  |  Google AI for Developers
"Introduced contextual caching for Gemini 1.5 Pro and Flash to optimize performance and cost for long prompts."