Business Document Generator Standards

Core Principles

"Template-First, Code-Second" We do NOT draw documents line-by-line with code. We inject data into pre-designed templates.

1. Separation of Concerns:
   • Design: Managed in .docx / .xlsx / .html files.
   • Logic: Managed in Python scripts using Golden Stack libraries.
2. Golden Tech Stack:
   • Word (.docx): docxtpl (Essential). use {{ jinja2_tags }} in Word.
   • Excel (.xlsx): openpyxl. For preserving existing styles/formulas.
   • PowerPoint (.pptx): python-pptx. Standard for slide generation.
   • PDF: WeasyPrint (HTML+CSS -> PDF). Best for styling and maintenance.
3. Output & Hygiene:
   • Output Path: All final files MUST be saved to docs/exports/.
   • Cleanup Policy: All temporary scripts and templates MUST be deleted after successful generation (rm script.py template.docx).
4. Korean Font Safety 🇰🇷:
   • Always strictly define fonts (e.g., NanumGothic) in CSS/Style to prevent □□□ (tofu) errors.

Quality Standards

• Validation: Generated files must be checked for existence and non-zero size.
• Context-Aware: The agent must clearly define the context dictionary before writing code.
• Dependency Check: Ensure libraries (docxtpl, openpyxl, weasyprint) are installed or prompted.

Phase 1: Design & Spec 📝

• Objective: Define the "Contract" between Template and Code.
• Action:
  1. Inspect User Request.
  2. Define Context Variables (e.g., user_name, total_revenue, item_list).
  3. Check if a template exists. If not, generate a "Base Template" creation script first.

Phase 2: Implementation 💻

• Word (docxtpl):

  from docxtpl import DocxTemplate
  doc = DocxTemplate("template.docx")
  context = { 'key': 'value' }
  doc.render(context)
  doc.save("output.docx")
  

• Excel (openpyxl):

  import openpyxl
  wb = openpyxl.load_workbook("template.xlsx")
  ws = wb.active
  ws['B2'] = "New Value"
  wb.save("output.xlsx")
  

• PDF (WeasyPrint):

  from weasyprint import HTML
  HTML(string=html_content).write_pdf("docs/exports/output.pdf")
  

• PowerPoint (python-pptx):

  [!IMPORTANT] Refer to `STANDARD_TEMPLATE_SPEC.md` for Layout Indices and Cleanup rules.

  from pptx import Presentation
  
  TEMPLATE_PATH = ".agent/references/biz-doc-generator/templates/standard_biz_template.pptx" 
  prs = Presentation(TEMPLATE_PATH)
  
  # [CRITICAL] Remove existing instructional slides SAFELY
  # Must iterate backwards and drop relationships to prevent file corruption
  if len(prs.slides) > 0:
      for i in range(len(prs.slides) - 1, -1, -1):
          rId = prs.slides._sldIdLst[i].rId
          prs.part.drop_rel(rId)
          del prs.slides._sldIdLst[i]
      
  prs.save("docs/exports/output.pptx")
  

Phase 3: Verification ✅

• Self-Check:
  • Did I use a template? (If I used add_paragraph loops, I failed).
  • Did I handle Korean fonts? (For PDF).
  • Is the output file saved correctly?

Checklist

• Stack Check: Am I using docxtpl / openpyxl / python-pptx / WeasyPrint?
• Path Check: Is the output pointing to docs/exports/?
• Cleanup: Did I schedule deletion of temp files?
• Template: Is there a template available or being created?
• Font: (PDF only) Is a Korean font explicitly specified in CSS?