What problem does it solve?
Academic paper generation workflows need a deterministic, sandboxed way to persist section drafts, assemble a manuscript, audit citations, and compile a real PDF without risking unsafe file access or corrupted artifacts.
Core Features & Use Cases
- Artifact Persistence & Assembly: Stores generated section fragments on disk and assembles them into a run-scoped LaTeX manuscript with a references.bib file under paper/<meta_run_id>/.
- Citation Auditing: Builds a citation provenance table classifying each reference as STRONG, OK, WEAK, INVALID, or UNUSED based on its URL, DOI, or arXiv locator.
- Hardened PDF Compilation: Runs xelatex and bibtex with shell escape disabled, paranoid Kpathsea file access, CJK line-breaking support, page-count verification, and bounded idempotent length-expansion repairs.
- Use Case: An orchestrator generating a research paper calls this runtime to persist seven section files, assemble paper.tex, verify every citation resolves to a credible source, and compile a page-count-verified PDF.
Quick Start
Invoke the runtime by piping a JSON payload with an operation field such as persist_sections, assemble_manuscript_tex, citation_map, or compile_pdf into scripts/run.py.
