# Migrating FinTally from React to Slint UI: Achieving a 100MB Native Memory Footprint

> Source: <https://anshugondi.hashnode.dev/migrating-fintally-from-react-to-slint-ui-achieving-a-100mb-native-memory-footprint>
> Published: 2026-11-02 05:41:25+00:00

**Recommendation:** For local-first apps that run on-device LLMs, compile the UI natively with a toolkit like Slint instead of shipping a web engine, then budget VRAM and RAM separately for active inference versus idle states.
More in the playbook "Inference and self-hosting": <https://wpnews.pro/playbooks/inference-serving>

# Migrating FinTally from React to Slint UI: Achieving a 100MB Native Memory Footprint

In local-first software architecture, system resources are non-negotiable. Web wrappers and heavy web engines tax system RAM before a single line of business logic executes.

To solve this, **FinTally** has officially completed its core frontend migration from **React.js to Slint UI** across three critical system endpoints:

1. **Auth**
2. **Budgets**
3. **Transactions**

By moving to statically compiled native UI bindings, the entire application interface now executes inside a sub-120MB memory envelope.

## 

Our testing baseline on a 6GB VRAM system yielded exceptional execution-on-demand specs. The table below outlines how local inference and idle execution scale across hardware tiers:

| Hardware Tier | Active AI / LLM Call Load | Idle / Post-Inference State | 
|---|---|---|
| **High-End GPUs***(e.g., RTX 5090 / 20GB VRAM Class)* | ~5 GB VRAM | ~100 MB System RAM *(Runs silently 24/7)* | 
| **Mid-Range GPUs***(6GB VRAM Baseline Test System)* | 3.4 GB VRAM & 500 MB – 600 MB RAM | 100 MB – 120 MB System RAM | 
| **Entry-Level GPUs***(4GB VRAM Class)* | ~2 GB VRAM | Sub-120 MB System RAM | 

## 

Because Slint UI compiles down to native machine instructions rather than relying on dynamic web interpreters, these low-footprint optimizations apply universally:

- **Mobile Compilation:** Light enough to execute natively on ARM mobile hardware once the entire route tree migration completes.
- **Desktop & WebAssembly (WASM):** Shared memory models guarantee high-throughput responsiveness regardless of execution target.

### 

We are actively working on pushing memory usage down even further, establishing a low-latency execution pipeline that seamlessly bridges local privacy-first inference with fallback cloud reliability.

*Follow the live engineering discussions on* *Twitter/X**and* *LinkedIn**.*

## About the author

- Systems Engineer focused on low-latency architectures, bare-metal C/C++ FFI, and production Rust backends. I specialize in eliminating runtime bottlenecks at the C-ABI boundary—building high-throughput polyglot extensions (PyO3, FFI) and zero-copy memory pipelines for AI/ML inference and SaaS infrastructure. Core Stack: Rust (Axum, Candle, SQLx), C++17/23, C23, PyO3, PostgreSQL, MinIO. Proof of Work: Creator of FinTally—a pure Rust finance platform evolved from legacy Node.js/Django to FastAPI + PyO3, and ultimately rewritten into a full-stack Rust architecture powered by bare-metal C/C++ native engines via extern "C" FFI.

## Comments

[Join the discussion](https://hashnode.com/posts/migrating-fintally-from-react-to-slint-ui-achieving-a-100mb-native-memory-footprint/6ae822857cfb115319e58df6)

No comments yet. [Be the first to comment.](https://hashnode.com/posts/migrating-fintally-from-react-to-slint-ui-achieving-a-100mb-native-memory-footprint/6ae822857cfb115319e58df6)
