Build the first higher-level authoring path for common Syntra bandit cases.\n\nGoal:\n- Accept a small YAML/JSON spec with name, options, contexts, algorithm, and reward coefficients.\n- Generate a Lycan source/capsule path suitable for installation into Syntra.\n- Keep Lycan available for advanced custom logic, but let common Syntra users start without writing .lycs by hand.\n\nInitial target:\n- LLM model router spec.\n- Compile or generate the equivalent of examples/demo_llm_model_router.lycs.\n- Document reward = weighted sum of normalized outcome metrics.\n\nThis is the main adoption unlock for Syntra 0.2.
Build the first higher-level authoring path for common Syntra bandit cases.\n\nGoal:\n- Accept a small YAML/JSON spec with name, options, contexts, algorithm, and reward coefficients.\n- Generate a Lycan source/capsule path suitable for installation into Syntra.\n- Keep Lycan available for advanced custom logic, but let common Syntra users start without writing .lycs by hand.\n\nInitial target:\n- LLM model router spec.\n- Compile or generate the equivalent of examples/demo_llm_model_router.lycs.\n- Document reward = weighted sum of normalized outcome metrics.\n\nThis is the main adoption unlock for Syntra 0.2.