Cloud? Nein. Your financial data stays local.

A mobile-first financial AI assistant for CFOs that runs entirely on-device using FunctionGemma + Cactus Compute. When cloud reasoning is needed, we use a novel reversible subgraph — anonymizing all entities before sending to Gemini, then de-anonymizing the response locally.

Real names never leave the device.

CFOs handle the most sensitive data in any company: SSNs, salaries, M&A details, cash positions, client revenue. They need instant answers between meetings, but they'd never paste this into ChatGPT. Current solutions force a choice: privacy or intelligence.

cloudNein uses a reversible subgraph — a bidirectional anonymization system:

1. Anonymize all entities (vendors, clients, employees, PII) into node aliases (Vendor_A, Client_B, Person_A)
2. Send anonymized structural data to Gemini for reasoning
3. De-anonymize the response locally — real names restored on-device

The cloud sees financial relationships without knowing who the actual entities are. The CFO gets specific, actionable advice with real names.

Reversible subgraph anonymization — standard redaction is one-way ("John Doe" → "[REDACTED]"), breaking AI reasoning. We preserve structure:

LOCAL DEVICE (never transmitted)    CLOUD (Gemini)
────────────────────────────────    ──────────────
Baker McKenzie  ↔  Vendor_A         Sees: Vendor_A
Acme Corp       ↔  Client_A         Sees: Client_A
Sarah Chen      ↔  Employee_A       Sees: Employee_A
123-45-6789     ↔  SSN_A            Sees: SSN_A

Gemini reasons over relationships: "Vendor_A has the highest legal spend. Consider renegotiating..."

Locally, we swap aliases back: "Baker McKenzie has the highest legal spend. Consider renegotiating..."

5-stage pipeline that autonomously decides what's safe to send to the cloud:

• Stage 0: PII detection + sensitivity scoring (<1ms)
• Stage 1: Complexity classification (data-lookup vs analytical)
• Stage 2: Context narrowing (7 tools → 2-3 relevant)
• Stage 3: FunctionGemma tool calling (~3s, on-device)
• Stage 4: Execution with 5 routing paths:
  • local-tool — SQL query on-device
  • cloud-tool — Gemini picks tool → SQL on-device
  • cloud-analysis — anonymize → Gemini → de-anonymize
  • privacy-redact — anonymize → Gemini compliance → de-anonymize
  • local-fallback — keyword extraction

• FunctionGemma 270M (via Cactus Compute) — on-device tool selection
• Gemini 2.5 Flash — cloud reasoning (only with anonymized data)
• React Native + Expo — cross-platform mobile
• SQLite — on-device financial database
• TypeScript — end-to-end type safety

Showstopper: Logs prove real vendor/client names never left the device.

npm install --legacy-peer-deps
npm run start

To make things work on your local simulator, or on your phone, you need first to run eas build. We have many shortcuts on package.json to make it easier:

npm run build:ios:sim # build for ios simulator
npm run build:ios:device # build for ios device
npm run build:ios:prod # build for ios device

This directory is designed to organize and store various assets, making it easy for you to manage and use them in your application. The assets are further categorized into subdirectories, including icons and images:

assets
├── icons
└── images

icons This is where your icon assets will live. These icons can be used for buttons, navigation elements, or any other UI components. The recommended format for icons is PNG, but other formats can be used as well.

Ignite comes with a built-in Icon component. You can find detailed usage instructions in the docs.

images This is where your images will live, such as background images, logos, or any other graphics. You can use various formats such as PNG, JPEG, or GIF for your images.

Another valuable built-in component within Ignite is the AutoImage component. You can find detailed usage instructions in the docs.

How to use your icon or image assets:

import { Image } from 'react-native';

const MyComponent = () => {
  return (
    <Image source={require('assets/images/my_image.png')} />
  );
};

Follow our Maestro Setup recipe.

Ignite Cookbook is an easy way for developers to browse and share code snippets (or “recipes”) that actually work.

Read our Upgrade Guide to learn how to upgrade your Ignite project.

⭐️ Help us out by starring on GitHub, filing bug reports in issues or ask questions.

💬 Join us on Slack to discuss.

📰 Make our Editor-in-chief happy by reading the React Native Newsletter.