Host-agnostic LLM agent execution on the OpenAI Agents SDK: model fallback, streaming, structured-output extraction, self-correction, and tracing — behind a small injected provider seam.
High-performance, memory-efficient, safe multimode sync <-> async channels built for real-time, low-overhead communication in concurrent Rust applications