Add APA DOCX manuscript export with reference deduplication

pyproject.toml

@@ -36,6 +36,7 @@ dependencies = [
     "nltk>=3.8",
     "langchain-text-splitters>=0.3.0",
     "pathspec>=1.1.1",
+    "python-docx>=1.1.0",
 ]
 
 [project.optional-dependencies]

src/perspicacite/rag/agentic/orchestrator.py

@@ -14,6 +14,7 @@
 from perspicacite.models.kb import chroma_collection_name_for_kb
 from perspicacite.provenance.context import get_collector
 from perspicacite.rag.dynamic_kb import DynamicKnowledgeBase
+from perspicacite.rag.export.apa_docx_exporter import export_apa_docx
 from perspicacite.rag.utils import format_references_academic
 from perspicacite.retrieval.hybrid import hybrid_retrieval
 
@@ -2427,6 +2428,15 @@ async def _generate_answer(
                 answer = answer.rstrip() + "\n\n" + references_section
                 logger.info("agentic_references_section_added", answer_chars=len(answer))
 
+        doc_id = session.session_id.split("-")[0] if session else "manuscript"
+        output_path = f"output/{doc_id}_manuscript.docx"
+        print("OUTPUT PATH:", output_path)
+        try:
+            export_apa_docx(answer, papers, output_path)
+            logger.info("agentic_manuscript_exported", path=output_path)
+        except Exception:
+            logger.warning("agentic_manuscript_export_failed", exc_info=True)
+
         return answer, citation_map
 
     _CITE_RE = re.compile(r"\[(\d+(?:\s*,\s*\d+)*)\]")
@@ -2626,6 +2636,15 @@ async def _generate_single_paper_answer(
         if references_section:
             answer = answer.rstrip() + "\n\n" + references_section
 
+        doc_id = session.session_id.split("-")[0] if session else "manuscript"
+        output_path = f"output/{doc_id}_manuscript.docx"
+        print("OUTPUT PATH:", output_path)
+        try:
+            export_apa_docx(answer, papers, output_path)
+            logger.info("agentic_manuscript_exported", path=output_path)
+        except Exception:
+            logger.warning("agentic_manuscript_export_failed", exc_info=True)
+
         return answer, citation_map
 
     def _build_numbered_paper_list(

src/perspicacite/rag/export/apa_docx_exporter.py

@@ -0,0 +1,56 @@
+from docx import Document
+from typing import List, Dict
+
+
+def format_authors(authors):
+    if not authors:
+        return ""
+
+    if len(authors) == 1:
+        return authors[0]
+
+    if len(authors) == 2:
+        return f"{authors[0]} & {authors[1]}"
+
+    return ", ".join(authors[:-1]) + f", & {authors[-1]}"
+
+
+def to_apa(paper: Dict) -> str:
+    authors = format_authors(paper.get("authors", []))
+    year = paper.get("year", "n.d.")
+    title = paper.get("title", "")
+    journal = paper.get("journal", "")
+    doi = paper.get("doi", "")
+
+    text = f"{authors} ({year}). {title}. {journal}."
+
+    if doi:
+        doi_clean = doi.replace("https://doi.org/", "")
+        text += f" https://doi.org/{doi_clean}"
+
+    return text
+
+
+def export_apa_docx(manuscript_text: str, papers: List[Dict], output_path: str):
+    doc = Document()
+
+    doc.add_heading("Manuscript", level=1)
+    doc.add_paragraph(manuscript_text)
+
+    doc.add_heading("References (APA Style)", level=1)
+
+    seen = set()
+    merged = []
+
+    for p in papers:
+        key = p.get("doi") or p.get("title")
+        if key in seen:
+            continue
+        seen.add(key)
+        merged.append(p)
+
+    for i, paper in enumerate(merged, 1):
+        doc.add_paragraph(f"{i}. {to_apa(paper)}")
+
+    doc.save(output_path)
+    return output_path