Building a Vertical Corpus Builder: Clean JSONL Datasets for LLM Fine-Tuning

Turning seed URLs into fine-tuning-ready corpora - boilerplate stripping, near-duplicate removal and token-aware JSONL chunking with full provenance.