I built a Rust client library for the Ollama API as a hands-on experiment with Rust and local LLM tooling, inspired by Il Codice di Marco YT videos and the Gemma 4 release.
Current scope: - async + blocking clients - chat / generate / embed - model operations - streaming responses
The implementation was handwritten, with AI used more as development support than code generation: - Claude Code (Opus 4.6) during development - GPT 5.4 for adversarial code review and robustness/security feedback
Any feedback is welcome.