Trampoline-AI/predict-rlm

**Brief Summary (2‑3 sentences)** Trampoline‑AI’s **predict‑rlm** is a production‑oriented runtime that lets a language model invoke sub‑models using DSPy‑style