Storia: Designing a Synthetic Data Pipeline for Persian LLM Fine Tuning: From Topic Graphs to QLoRA Evaluation — Warptech Lab News