What problem does it solve?
It produces a final delivery summary for an automated paper-writing workflow without relying on an LLM, ensuring that reported PDF page counts and citation statistics come only from verified, machine-readable upstream outputs rather than inferred prose.
Core Features & Use Cases
- Verified PDF Reporting: Parses PDF_PATH, PDF_PAGES, and PDF_TARGET_PAGES markers from compile output and fails closed if they are missing, duplicated, or inconsistent.
- Citation Summary Validation: Requires a complete, internally consistent citation SUMMARY line (total, strong, ok, weak, invalid, unused) before rendering any statistics.
- Bilingual Output: Renders the delivery note in Chinese or English based on the confirmed paper language contract, with a neutral fallback for other languages.
- Use Case: After an automated paper-writing pipeline compiles a PDF and maps citations, this formatter emits the final user-facing delivery message with exact page counts, citation quality breakdown, and warnings about weak or unused references.
Quick Start
Run the render script with a JSON payload containing the paper contract, language instruction, compile_pdf output, and citation_map output to print the verified delivery summary.