Best Practices Advanced

Running AI workloads on Firebase at scale requires careful attention to costs, security, model selection, and monitoring. This lesson covers the essential patterns for production Firebase AI applications.

Cost Management

Strategy Implementation Impact
Use Flash models Choose gemini-2.0-flash over gemini-2.0-pro for simple tasks 10x cost reduction
Cache AI responses Store results in Firestore for identical or similar queries 50-90% savings on repeat queries
Set billing alerts Configure budget alerts in Google Cloud Console Prevents unexpected bills
Limit max tokens Set maxOutputTokens on all model calls 20-50% savings
Batch embeddings Generate embeddings in bulk, not one at a time Reduces function invocations

Security Rules for AI

Firestore Rules 
rules_version = '2';
service cloud.firestore {
  match /databases/{database}/documents {

    // AI-generated content: users can only read their own
    match /ai-responses/{responseId} {
      allow read: if request.auth != null
        && resource.data.userId == request.auth.uid;
      allow create: if false; // Only Cloud Functions can write
    }

    // Rate limit: max 10 AI requests per minute per user
    match /ai-requests/{requestId} {
      allow create: if request.auth != null
        && request.resource.data.userId == request.auth.uid;
    }
  }
}

Model Selection Guide

Task Recommended Model Why
Simple Q&A, classification Gemini 2.0 Flash Fastest, cheapest, good enough
Complex reasoning, analysis Gemini 2.0 Pro Better accuracy for hard tasks
Image/video understanding Gemini 2.0 Flash (multimodal) Built-in vision, fast processing
Text embeddings text-embedding-004 High quality, reasonable cost
On-device OCR/detection ML Kit Free, instant, offline

Enable App Check

Protect your AI endpoints from abuse:

TypeScript 
import { initializeAppCheck, ReCaptchaV3Provider } from 'firebase/app-check';

// Enable App Check to verify legitimate requests
const appCheck = initializeAppCheck(app, {
  provider: new ReCaptchaV3Provider('your-recaptcha-site-key'),
  isTokenAutoRefreshEnabled: true,
});

Monitoring Checklist

Monitor These Metrics:
  • Cloud Function execution time and error rate
  • Vertex AI API usage and costs per day
  • Firestore reads/writes from AI functions
  • App Check rejection rate (indicates abuse attempts)
  • Model response latency (p50, p95)
  • User-reported AI quality issues

Production Architecture

Text 
Client App                Firebase                    Google Cloud
+------------------+     +------------------+        +------------------+
| Firebase Auth    | --> | App Check        | -----> | Vertex AI        |
| Vertex AI SDK    | --> | Firestore        |        | Gemini Models    |
| ML Kit (device)  |     | Cloud Functions  | -----> | Embeddings API   |
+------------------+     | Cloud Storage    |        | Vector Search    |
                          +------------------+        +------------------+

Course Complete!

Congratulations! You have completed the Firebase + AI course. You can now build intelligent applications using Firebase with Vertex AI, Gemini, ML Kit, Cloud Functions, and Firestore vector search.

← Back to Course Overview
← Cloud Functions Course Overview →

Ready to Go Deeper?

Live instructor-led courses from our partners. Affiliate disclosure.

🧠

AI & ML Courses - 30% Off

Live instructor-led AI, machine learning, data science, and cloud courses for working professionals. Use code Limited30 at checkout.

Edureka
📊

DataCamp - AI & Data Science

Hands-on Python, machine learning, and AI courses with interactive exercises and real projects.

DataCamp
🎓

edX - Top AI Courses

University-level AI courses from MIT, Harvard, Stanford. Earn certificates that employers recognize.

edX