Firebase is the classic serverless backend. Cencori adds the AI layer that Firebase lacks.

Cloud Functions (v2)

The most common pattern is to wrap Cencori calls in a compiled Cloud Function to keep your API key secure.

Codetext

// functions/src/index.ts
import { onRequest } from "firebase-functions/v2/https";
import * as logger from "firebase-functions/logger";
import { Cencori } from "cencori";
 
const cencori = new Cencori({
  apiKey: process.env.CENCORI_API_KEY,
});
 
export const chat = onRequest(async (request, response) => {
  // 1. App Check / Auth Validation
  // ... verify token ...
 
  // 2. Execute AI logic
  try {
    const result = await cencori.chat.create({
        model: "gpt-4o",
        messages: request.body.messages,
        exclude_pii: true // Enable Cencori Security
    });
 
    response.json(result);
  } catch (error) {
    logger.error("AI Error", error);
    response.status(500).send("AI Inference Failed");
  }
});

Firestore Triggers

You can create "background agents" that react to database changes.

Example: A user writes a note, and Cencori automatically tags it.

Codetext

import { onDocumentCreated } from "firebase-functions/v2/firestore";
 
export const tagNote = onDocumentCreated("notes/{noteId}", async (event) => {
  const note = event.data?.data();
  if (!note) return;
 
  // Ask Cencori to generate tags
  const tags = await cencori.utils.tag({
    content: note.text,
    categories: ["Work", "Personal", "Code", "Ideas"]
  });
 
  // Write tags back to Firestore
  return event.data.ref.set({ tags }, { merge: true });
});

Security Rules

Since Cencori handles the AI logic, you can keep your security rules simple.

Codetext

// firestore.rules
service cloud.firestore {
  match /databases/{database}/documents {
    match /notes/{note} {
      // Only allow users to read/write their own notes
      allow read, write: if request.auth.uid == resource.data.userId;
    }
  }
}