Top AI Repos — open-source AI, indexed and scored
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Top AI Repos tracks AI repositories on GitHub and answers two different questions about each one: is it moving right now, and would you bet a product on it.
Lightweight LLM firewall: masks PII, routes calls, leaves zero trace.
| Date | Stars |
|---|---|
| 2026-07-31 | 377 |
| 2026-08-06 | 377 |
Today
— stars today
This week
— stars this week
This month
— stars this month
Momentum
0.0
growth rate 0.00%/day
OneAIFW === OneAIFW is a local, lightweight “AI firewall” that anonymizes sensitive data before sending it to LLMs, and restores it after responses. - Core engine: Zig + Rust (WASM and native) - JS binding: libs/aifw-js (Transformers.js + WASM aifw core) - Python binding: libs/aifw-py (Python's transformers package + native aifw core) - Demos: Web app(apps/webapp), Browser Extension(browser_extension) - Backend service based on presidio/LiteLLM: `py-origin` (FastAPI + Presidio/LiteLLM) ## What we protect for you Privacy: - Physical Address - Email Address - Name - Phone - Bank Account - Paymant Information Secrets: - Verification Code - Password Crypto: - Seed - Private Key - Address ## The Demo website The site at [oneaifw.com](https://oneaifw.com/) is a full end‑to‑end demo of the OneAIFW project. It showcases how OneAIFW detects sensitive information, anonymizes it before LLM calls, and restores it afterward, all in a browser‑friendly UI backed by the same core engine used in this repository. ## Monorepo layout - `core/` Zig core and pipelines (mask/restore) - `libs/aifw-js/` JavaScript library used in browser and demos - `apps/webapp/` Minimal browser demo (Vite) - `browser_extension/` Chrome/Edge extension sample - `py-origin/` Python backend service + CLI command (see its own README) - `tests/transformer-js/` Transformers.js demo and model prep scripts ## Architecture High‑level architecture (overview): OneAIFW is built as a layered, cross‑platform stack around a single core library. The aifw core library is designed to compile both as a native library (for backend services and CLIs) and as a WASM module (for browsers and JS runtimes), with language bindings providing ergonomic APIs on top. - **Core engine: aifw core library (Zig + Rust)** Cross‑platform engine that implements the masking/restoring pipelines and regex/NER span fusion. It builds to: - Native libraries for use from Python and other host languages. - WASM (`wasm32-freestanding`) for use in browsers and JS environments. - **Language bindings** - **aifw-js (`@oneaifw/aifw-js`)**: JavaScript/TypeScript binding that runs NER with Transformers.js, converts spans to byte offsets, calls the WASM core to mask/restore, and exposes a high‑level API (including batch, language‑aware masking, and integration helpers for web apps and extensions). - **aifw-py**: Python binding that loads the native core library and exposes a simple API (`mask_text`, `restore_text`, batch variants, and configuration) used by Python CLIs and HTTP services. - **Backends and apps** - **Web demo (`apps/webapp`)**: Vite‑based frontend that talks to the core via `aifw-js`, demonstrating end‑to‑end masking/restoring flows entirely in the browser or against a backend service. - **Browser extension (`browser_extension`)**: Example Chrome/Edge extension that injects OneAIFW into arbitrary pages to protect prompts before they are sent to LLM UIs. - **Python services (`cli/python`)**: HTTP APIs and CLI wrappers built on top of `aifw-py`, suitable for running as local daemons or containerized services behind gateways. - **Presidio based services (`py-origin`)**: HTTP APIs built on top of presidio library, suitable for running as local daemons or containerized services behind gateways. - **Production demo website** The public demo at [oneaifw.com](https://oneaifw.com/) is built from this stack and uses the same core engine and bindings described above. ## Prerequisites - Zig 0.15.2 - Rust toolchain (stable) + Cargo - `rustup target add wasm32-unknown-unknown` - Node.js 18+ and pnpm 9+ - Install pnpm: `npm i -g pnpm` - Python 3.10+ (for `py-origin` backend) and pip/venv Verify versions: ```bash zig version # expect 0.15.2 rustc --version cargo --version node -v pnpm -v python3 --version ``` ## Getting Started OneAIFW lets you safely call external LLM providers by anonymizing sensitive data first, then restoring it after the mod
Excerpt of 14,027 characters
Read on GitHubWould you bet a product on this? Bounded 0–100 and slow moving.
matched fp:6fc6b131a2709e86, llm:description: 'Lightweight LLM firewall: masks PII, routes calls, leaves zero trace.' (repository description)
matched fp:6fc6b131a2709e86, llm:description: 'Lightweight LLM firewall: masks PII, routes calls, leaves zero trace.' (repository description)
matched fp:6fc6b131a2709e86, llm:description: 'Lightweight LLM firewall: masks PII, routes calls, leaves zero trace.' (repository description)