| # Step 2 Outputs | |
| - `chunks.jsonl`: validated leaf chunks for each document, including `graph_edges` with cross-reference metadata | |
| - `chunks.md`: full readable Markdown export of the chunk outputs | |
| - `summary.json`: per-document counts, parser routes, graph-edge counts, and validation results | |
| Step 2 now chunks from step-1 structured artifacts: | |
| - `raw.md` for legal structure | |
| - `clean_pages.jsonl` for administrative narrative flow | |
| - `raw.json` for tables and appendices | |