CLAUSULA_AI：基于LangGraph多智能体的巴西法律合同AI审查平台

CLAUSULA_AI is an intelligent contract analysis and management platform tailored for Brazilian legal and accounting firms. It leverages LangGraph multi-agent workflows to automate contract review, supporting full-process functions like DREI compliance detection, clause rewriting, and electronic signatures. The platform uses a three-layer architecture (React Web dashboard + FastAPI backend + Word plugin) to integrate AI capabilities directly into lawyers' daily work environments, improving review accuracy and efficiency.

In the legal industry, contract review is time-consuming and error-prone. For Brazilian firms, ensuring compliance with DREI (Department of Business Registration and Integration) regulations via manual review is inefficient and risky. CLAUSULA_AI addresses these issues by automating the process, reducing human error and saving time.

Three-layer Architecture:

1. Microsoft Word Add-in (Office.js, React/TS/Vite)
2. Web Dashboard (React/TS/Vite, Zustand, TipTap)
3. FastAPI Backend (Python3.11, Uvicorn) connected to LangGraph Agents (GPT-4o), Firebase (auth/storage), and ZapSign (e-signature).

Tech Stack Details:

• Backend: FastAPI, LangGraph+LangChain+GPT-4o, PyMuPDF (doc parsing), Tesseract (OCR), Firebase Admin, ZapSign API.
• Frontend: React18+TS, Vite5, Zustand, TipTap, Tailwind CSS.
• Word Plugin: Office.js, React18+TS, Vite5, Tailwind CSS.

The platform uses a four-node LangGraph pipeline:

1. CleanTextNode: Removes OCR artifacts, page numbers, and noise from raw text.
2. DetectIssuesNode: GPT-4o-driven audit for DREI compliance, CNAE alignment, capital clause clarity, and accounting governance. Returns structured issues with references.
3. ClassifyIssuesNode: Maps issues to PDF bounding boxes and standardizes severity (deterministic, no LLM call).
4. GenerateSummaryNode: Creates structured review summaries with issues, suggestions, and compliance assessments.

• Smart Contract Review: Auto-detects compliance defects, ambiguous clauses, and logical contradictions with precise location marking.
• Spatial Highlighting: Links AI-detected issues to PDF paragraph bounding boxes for visual highlighting.
• AI-assisted Clause Rewriting: Generates DREI-compliant alternatives for problematic clauses (lawyers can accept/modify/keep original).
• Multi-format Export: PDF (ReportLab), Word (python-docx), Excel (openpyxl) reports.
• Cloud Storage: Firebase Storage for secure document storage with version control.
• E-signature: ZapSign API integration for end-to-end signing workflows.
• Word Plugin: Embeds AI analysis directly into Word, enabling seamless workflow integration.

Docker Compose Deployment: Run docker compose up --build to start the stack. Access:

• Frontend Dashboard: http://localhost:5173
• Backend API: http://localhost:8000
• Swagger Docs: http://localhost:8000/api/docs

Environment Variables:

• Backend: OPENAI_API_KEY, FIREBASE_CREDENTIALS_PATH, ZAPSIGN_API_TOKEN.
• Frontend: VITE_API_URL, Firebase configs.

• Legal Firms: Reduces initial contract review time from hours to minutes, letting lawyers focus on complex tasks.
• Accounting Firms: Ensures client documents (charters, shareholder agreements) comply with DREI regulations, lowering compliance risks.
• In-house Legal Departments: Standardizes review processes, improves consistency, and reduces human errors.

CLAUSULA_AI demonstrates AI's potential in legal tech. It uses LangGraph multi-agent architecture to automate contract review and integrates AI into Word for seamless use. The platform is open-source (hosted on GitHub) and welcomes community contributions: new contract types, compliance rules for other jurisdictions, UI/UX improvements, and performance optimizations. Its value lies in augmenting (not replacing) legal professionals, freeing them from repetitive work to focus on high-value tasks.