Python/ Streamlit - AI-Powered Academic Word/ Document Generator We are looking for an experienced Python developer to build a Streamlit web application that generates fully formatted academic Word documents (.docx) using multiple AI providers (OpenAI, Anthropic Claude, Google Gemini). The system must produce a complete word documents: cover page, abstract, automatic TOC, chapters, real Word footnotes, and bibliography. The user fills in a form (document type, title, language, page count, citation style, chapter structure, etc.), optionally uploads guideline files (PDF/DOCX/TXT), and the app calls an AI API to generate a complete document—cover page, abstract, table of contents, chapters, real Word footnotes, and bibliography—saved as a properly formatted .docx file ready for academic submission. The primary output language is Romanian, but the system must also support any other language. Generated documents must comply with Romanian academic formatting standards (Times New Roman 12pt, 1.5 spacing, specific margins, proper heading hierarchy, etc.). You are free to design your own architecture. A reference prototype is available to help you understand the expected output. Core Technical Requirements • Advanced .docx Engine (Non-Negotiable): o Native Word Footnotes: You must use python-docx with low-level OXML access to create native w:footnote elements in footnotes.xml. Simple superscript text is NOT acceptable. o Word-Native TOC: Implementation of field codes that allow the user to "Update Table" directly in Microsoft Word. o Academic Formatting: Strict compliance with standards: Times New Roman 12pt, 1.5 spacing, 3/2/2.5/2.5 cm margins, and justified text. • Romanian Language & Diacritics: o Full support for Romanian diacritics (ă, â, î, ș, ț) is mandatory. You must ensure encoding (UTF-8) is handled correctly in AI prompts, text extraction, and the final .docx output. • Multi-AI Pipeline: A robust fallback system (OpenAI → Claude → Gemini) to handle rate limits or API downtime mid-generation. • Guideline Injection: Extract text from uploaded PDF/DOCX guidelines and inject it into the AI prompt as authoritative formatting/style rules. • Registry & Editing: A JSON-based document registry and a dedicated UI page to modify/regenerate specific sections of existing documents. Additional Info • Payment: 100% Escrow – released only after the project is fully functional on a Windows environment and meets all requirements. • Detailed technical specifications is available on request (separate file). Please review them before applying. • A reference prototype is available on request (for inspiration only – not required as a starting point). • PLEASE start your application with the phrase: 'I am a python-docx expert'. If this is missing, I will archive your proposal without reading it. • Proposals that include a brief technical plan (2–3 sentences on architecture and key libraries) will be ranked first. • Please briefly explain how you plan to handle the footnotes.xml relationship within the Word document package to ensure that footnotes are recognized as native objects by Microsoft Word.