{"slug": "why-we-are-replacing-react-and-model-proxies-with-bare-metal-c23-rust-and-slint", "title": "Why We are Replacing React and Model Proxies with Bare-Metal C23, Rust, and Slint UI in FinTally", "summary": "A developer rebuilt FinTally's core engine around a bare-metal C23 and Rust voice subsystem, replacing React and continuous model inference with a bounded-memory pipeline that uses 8–40 MB of working memory and 10–40 µs of FFI overhead. The system runs deterministic MFCC, zero-crossing rate, energy and pitch extraction in C23, a C++17 vision engine for invoice binarization, and lock-free SPSC ring buffers in Rust, with LLMs limited to drafting review prompts while money-moving actions require human approval. The frontend moved to Slint UI, which cross-compiles to desktop, mobile and WebAssembly.", "body_md": "# Why We are Replacing React and Model Proxies with Bare-Metal C23, Rust, and Slint UI in FinTally\n\nBuilding an edge-first, fail-closed voice & vision pipeline with ~8-40 MB RAM footprints and sub-millisecond local triggers.\n\nWhen building financial software, \"make it fast\" isn't a strategy—and letting unvalidated AI models hold execution authority is a disaster waiting to happen.\n\nOver the past few weeks, we overhauled the core engine of FinTally. We moved away from web-heavy runtimes (React JS) and expensive continuous model inference, pivoting to a bare-metal, native-first architecture.\n\n### \n\nRunning deep learning speech models like Whisper 24/7 just to listen for voice triggers burns continuous CPU cycles, requires 500MB–2GB+ of VRAM, and adds 200ms+ of latency per audio chunk.\n\nInstead, we built `fintally_voice`: a bounded-memory C23 + Rust subsystem that acts as a microsecond gatekeeper.\n\n### \n\n1. Deterministic Native Kernels (C23 & C++17)\n\n- C23 Core (~8-40 MB working set): Handles MFCC extraction, Zero-Crossing Rate (ZCR), energy levels, and F0 pitch tracking.\n- C++17 Vision Engine: Binarizes invoices and extracts spatial layout deterministically before any text hits an LLM.\n\n1. Safe Rust Runtime & Zero-Copy FFI\n\n- Lock-free Single-Producer Single-Consumer (SPSC) ring buffers handle frame passing with only 10–40 µs FFI overhead.\n- Strict financial safety rule: LLMs only draft review prompts. Money-moving actions require explicit human approval and fail-closed validation.\n\n1. Slint UI: Cross-Platform Without JS Taxes We purged the React frontend. Using Slint UI, a single codebase now cross-compiles natively to Desktop, Mobile, and WebAssembly (WASM) with sub-second boot times and near-zero memory footprint.\n\n🔗 **Cross-Platform Links:**\n\n- Follow the discussion thread on X: [https://x.com/AnshuGondi/status/2113835272186347712](https://x.com/AnshuGondi/status/2113835272186347712)\n- Check out the full dev branch on GitHub: [https://github.com/Anshu-Gondi/FINTALLY-Finance-Monitoring-app/tree/dev](https://github.com/Anshu-Gondi/FINTALLY-Finance-Monitoring-app/tree/dev)\n\n## About the author\n\n- 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.\n\n## Comments\n\n[Join the discussion](https://hashnode.com/posts/why-we-are-replacing-react-and-model-proxies-with-bare-metal-c23-rust-and-slint-ui-in-fintally/6adc216b2811ca40f2ba8e27)\n\nNo comments yet. [Be the first to comment.](https://hashnode.com/posts/why-we-are-replacing-react-and-model-proxies-with-bare-metal-c23-rust-and-slint-ui-in-fintally/6adc216b2811ca40f2ba8e27)", "url": "https://wpnews.pro/news/why-we-are-replacing-react-and-model-proxies-with-bare-metal-c23-rust-and-slint", "canonical_source": "https://anshugondi.hashnode.dev/why-we-are-replacing-react-and-model-proxies-with-bare-metal-c23-rust-and-slint-ui-in-fintally", "published_at": "2026-10-24 03:09:31+00:00", "updated_at": "2026-11-21 06:14:36.824605+00:00", "lang": "en", "topics": ["ai-agents", "ai-infrastructure", "developer-tools", "ai-tools"], "entities": ["FinTally", "Slint UI", "Rust", "React", "Whisper", "Anshu Gondi", "GitHub"], "also_reported_by": [], "takeaway": "Gating voice triggers with deterministic C23 signal features (MFCC, ZCR, energy, F0) instead of running Whisper continuously cuts the working set to 8–40 MB and FFI frame-passing overhead to 10–40 µs.", "recommendation": {"advice": "Put a cheap deterministic gate in front of expensive model inference and keep LLM output advisory: require explicit human approval and fail-closed validation before any money-moving action executes.", "playbook": "agent-architecture", "playbook_url": "https://wpnews.pro/playbooks/agent-architecture"}, "alternates": {"html": "https://wpnews.pro/news/why-we-are-replacing-react-and-model-proxies-with-bare-metal-c23-rust-and-slint", "markdown": "https://wpnews.pro/news/why-we-are-replacing-react-and-model-proxies-with-bare-metal-c23-rust-and-slint.md", "text": "https://wpnews.pro/news/why-we-are-replacing-react-and-model-proxies-with-bare-metal-c23-rust-and-slint.txt", "jsonld": "https://wpnews.pro/news/why-we-are-replacing-react-and-model-proxies-with-bare-metal-c23-rust-and-slint.jsonld"}}