Stripping Out Python and Node.js: How Migrating Kinetib to Rust, C23, and Slint UI Cut Idle RAM Below 120 MB A developer behind Kinetib, a local-first personal finance and AI assistant platform, migrated the stack from Node.js, Django, and FastAPI to a native engine built on Rust (Axum/Tokio), C23, C++17, and Slint UI, using PyO3 and C++ N-API addons as a transitional bridge. The rewrite runs quantized Qwen2.5 GGUF and bge-small-en-v1.5 embeddings via Candle and USearch HNSW, with native Tesseract OCR and a lock-free C23 acoustic DSP core. Stripping the interpreted runtimes cut idle memory to under 120 MB, even before Slint UI tuning is complete, according to the developer. Stripping Out Python and Node.js: How Migrating Kinetib to Rust, C23, and Slint UI Cut Idle RAM Below 120 MB Most AI apps demand gigabytes of RAM and cloud GPUs just to answer a basic question. When I started building Kinetib —a local-first personal finance and AI assistant platform—I originally built it purely as a web application. At the time, the architecture looked like a typical modern AI stack: Node.js, Django, and FastAPI services orchestrating local model runners experimenting with TinyLlama, Florence OCR, Qwen2.5 GGUF, and bge-small-en-v1.5 . It worked, but carrying multiple interpreted runtimes and web wrappers created massive memory overhead and startup latency. To deliver a truly private, zero-cloud-lag experience on everyday hardware, I made the decision to pivot away from a web-only architecture and systematically migrate the entire stack into pure systems programming. | Layer | Legacy Web Stack Before | Unified Native Engine After | |---|---|---| | Frontend & Target | Web-only React.js | Slint UI Targeting Desktop, Mobile, Tablet & Wasm/Cloud from one codebase | | Backend & Core | Node.js, Django, FastAPI | Rust Axum/Tokio + extern "C" FFI boundaries | | Migration Bridge | Python/Node wrappers | Phased out via PyO3 and C++ N-API addons during transition | | AI & Embeddings | Multi-model Python runners | Quantized Qwen2.5 GGUF + bge-small-en-v1.5 via Candle & USearch HNSW | | OCR & Voice DSP | Heavy Python vision/audio wrappers | Native C++17/Rust Tesseract OCR + lock-free C23 acoustic DSP core | | Idle RAM Footprint | Multi-runtime bloat GBs under load | < 120 MB RAM even before finishing Slint UI tuning | Right now, I am in the middle of a 20–30 day rewrite porting the remaining frontend endpoints to Slint UI so Kinetib can natively target desktops, mobiles, tablets, and web/cloud deployments simultaneously without sacrificing local-first execution. What surprised me most during benchmarks is that even with my Slint UI layer still largely unoptimized on my end, stripping out the interpreted runtimes and Python wrappers immediately dropped the idle memory footprint to under 120 MB of RAM . By replacing heavy Python wrappers with a C23 acoustic DSP static library, a C++17/Rust OCR pipeline, and pure Rust orchestration, every millisecond of latency and megabyte of memory is accounted for. Users shouldn't have to upload their bank transactions, budgets, or EMI schedules to a third-party cloud server just to get fast AI insights. When you eliminate cloud inference bills and runtime bloat, you can ship software that runs instantaneously on constrained hardware while keeping user data 100% private. You can inspect the core workspace and migration architecture under the Apache 2.0 license on my GitHub Anshu-Gondi/Kinetib-Finance-Monitoring-app . Follow the ongoing build updates on LinkedIn and X Twitter . 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/stripping-out-python-and-node-js-how-migrating-kinetib-to-rust-c23-and-slint-ui-cut-idle-ram-below-120-mb/6b012966ddb44ab13ecee46e No comments yet. Be the first to comment. https://hashnode.com/posts/stripping-out-python-and-node-js-how-migrating-kinetib-to-rust-c23-and-slint-ui-cut-idle-ram-below-120-mb/6b012966ddb44ab13ecee46e