@eloqnt/inference

The eloqnt translation pipeline: prompt assembly, model inference, and quality gates — shared by the SDK (runs locally) and the engine (runs hosted)