# 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