Show HN: Ouverture.py – Content-addressed storage for multilingual functions (github.com)

Ouverture.py is a research prototype that builds a content‑addressed “function pool” for multilingual Python code: logically identical functions written with different human-language identifiers (English, French, Spanish, etc.) normalize to the same hash and are stored together so each language’s version can be reconstructed on demand. The tool parses Python into an AST, extracts docstrings, canonicalizes variable names (e.g., _ouverture_v_0), computes a SHA‑256 on the normalized AST (excluding docstrings), and stores language-specific name mappings using ISO 639‑3 codes in .ouverture/objects/XX/YYYYYY.json. Imports are preserved, inter-function references are normalized to hash-based imports (from ouverture import <HASH>), and retrieval can reconstruct a function in a target language. It requires Python 3.9+ (ast.unparse) and is MIT‑licensed. This matters for AI/ML and developer communities because it tackles linguistic bias in code: by separating logic from natural‑language surface choices, Ouverture can preserve cognitive diversity, make code reuse language‑agnostic for humans and LLMs, and provide canonical forms that may improve multilingual code search and LLM training/interaction. It’s purely syntactic (no semantic analysis), limited to Python, known to have bugs and edge cases, and explicitly not a blockchain project. The repo is experimental and welcomes non‑English examples, bug reports, and alternative implementations to explore whether native‑language programming improves comprehension and model performance in a post‑LLM world.