README.md

metadata

title: ClauseGuard
emoji: 🛡️
colorFrom: gray
colorTo: gray
sdk: gradio
sdk_version: 5.23.0
python_version: '3.12'
app_file: app.py
pinned: false

🛡️ ClauseGuard v4.3 — World's Best Open-Source Legal Contract Analysis

ClauseGuard is the most comprehensive open-source AI-powered legal contract analysis tool. It analyzes contracts using state-of-the-art legal NLP models and provides actionable risk assessments, Q&A chatbot, clause redlining, and OCR for scanned PDFs.

🆕 What's New in v4.3

Feature	Description
⚡ ONNX + INT8 Quantization	CUAD classifier now supports ONNX Runtime with dynamic INT8 quantization — 2-4x faster inference on CPU. New ml/export_onnx_v2.py handles the full merge→export→quantize pipeline.
🎯 Better Embeddings	Upgraded from all-MiniLM-L6-v2 to BAAI/bge-small-en-v1.5 — +21% retrieval accuracy on MTEB benchmarks, same 384-dim, same latency. Includes query instruction prefix for asymmetric retrieval.
🚀 Batched Classification	All clauses classified in a single batched forward pass (batch_size=8) instead of one-by-one — 2-3x throughput improvement.
🧵 CPU Thread Control	torch.set_num_threads(2) prevents CPU thrashing under concurrent Gradio requests

Previous: v4.2

Feature	Description
🔧 NLI Fix	Fixed contradiction detection — now uses CrossEncoder.predict() instead of broken pipeline("text-classification") dict input. Contradictions actually work now.
🔒 Thread Safety	BoundedCache now uses threading.RLock to prevent race conditions under concurrent Gradio requests
⚡ Pre-compiled Regex	All regex patterns (clause classification, obligations, compliance negation) pre-compiled at module level — eliminates thousands of redundant compilations
🔗 Extension Fix	Chrome extension risk formula now matches backend (diminishing returns, not normalized by doc length). Fixed API_BASE URL.
🏷️ Label Coverage	Added missing regex-only labels (Indemnification, Confidentiality, Force Majeure, Penalties) to RISK_MAP and DESC_MAP
🛡️ Security	API CORS localhost origins now require explicit opt-in via CORS_ALLOW_LOCALHOST=true env var

Previous: v4.0

Feature	Description
🔍 OCR for Scanned PDFs	Smart PDF router: auto-detects native vs scanned PDFs. Scanned PDFs are processed via docTR OCR engine (CPU-friendly, ~150MB models)
💬 Contract Q&A Chatbot	RAG-powered chatbot that answers questions about your analyzed contract. Uses sentence-transformers for retrieval + Qwen2.5-7B via HF Inference API for generation
✏️ Clause Redlining	3-tier system: (1) Template lookup from 18+ legal templates based on FTC/EU standards, (2) Keyword-based matching, (3) LLM refinement for CRITICAL/HIGH risk clauses

✨ Core Features

Analysis Engine

Feature	Description
41 CUAD Clause Categories	Full taxonomy: Document Name, Parties, Governing Law, Indemnification, Termination, Non-Compete, IP Ownership, Audit Rights, Force Majeure, and more
4-Tier Risk Scoring	Critical 🔴 / High 🟠 / Medium 🟡 / Low 🟢 with visual risk matrix
Legal NER	Extracts parties, dates, monetary values ($), jurisdictions, defined terms, and party roles
NLI Contradiction Detection	Identifies conflicting clauses (e.g., uncapped + capped liability) and missing critical provisions
Obligation Tracker	Categorizes action items: monetary 💰, compliance ⚖️, reporting 📊, delivery 📦, termination 🛑
Compliance Checker	Validates against GDPR, CCPA, SOX, HIPAA, and FINRA requirements
Contract Comparison	Side-by-side diff between two contracts with alignment scoring
Clause Redlining	Suggests safer alternatives for risky clauses with legal citations
Q&A Chatbot	Ask questions about your contract using RAG (Retrieval-Augmented Generation)
OCR Support	Process scanned PDFs with docTR OCR engine

Document Support

• PDF parsing via pdfplumber (native) + docTR OCR (scanned)
• DOCX/DOC parsing via python-docx
• TXT / Markdown direct text input

UI/UX

• 3-Panel Professional Layout — Upload sidebar + Main analysis + Summary dashboard
• Document Viewer — Inline entity highlights (colored annotations)
• Clause Cards — Expandable risk-badged cards with confidence scores
• Redlining Tab — Side-by-side original vs suggested safer alternatives
• Q&A Chat Tab — Conversational interface to ask questions about the contract
• Export Reports — JSON (structured) and CSV (tabular) downloads
• Color-Coded Risk Badges — Instant visual triage

🧠 Models & Architecture

Component	Technology
Clause Classification	Mokshith31/legalbert-contract-clause-classification — LoRA adapter on nlpaueb/legal-bert-base-uncased, fine-tuned on CUAD 41-class taxonomy
Legal NER	matterstack/legal-bert-ner (ML) with regex fallback for 7 entity types
NLI	cross-encoder/nli-deberta-v3-base (semantic contradiction detection)
Embeddings	BAAI/bge-small-en-v1.5 (384-dim, RAG retrieval — +21% over MiniLM)
LLM	Qwen/Qwen2.5-7B-Instruct via HF Inference API (chatbot + redlining)
OCR	docTR (fast_base + crnn_vgg16_bn) for scanned PDF text extraction
Compliance	Regulatory keyword matching across GDPR, CCPA, SOX, HIPAA, FINRA
Comparison	Semantic similarity with sentence embeddings + string matching fallback
Obligations	Regex pattern matching across 5 obligation categories

🔍 OCR Architecture (Smart PDF Router)

PDF uploaded
    ↓
[detect_if_scanned] — pdfplumber extracts >50 chars/page?
    ↓                           ↓
  Native PDF               Scanned PDF
    ↓                           ↓
  pdfplumber              docTR OCR (CPU)
    ↓                           ↓
  Contract text → existing analysis pipeline

💬 Q&A Chatbot Architecture (RAG)

User asks question about their contract
        ↓
[1] Embed question with all-MiniLM-L6-v2
        ↓
[2] Retrieve top-5 most relevant chunks from contract
        ↓
[3] Build prompt:
    - System: ClauseGuard analysis results (clauses, entities, risk scores)
    - Context: Retrieved contract chunks (≤2.5K tokens)
    - User question
        ↓
[4] Stream response from Qwen2.5-7B via HF Inference API

Key design: Analyzed data (clauses, entities, risk scores) goes in the system prompt — NOT through RAG retrieval. Only the raw contract text goes through RAG. This gives the model both structured analysis AND verbatim evidence.

✏️ Clause Redlining Architecture (3-Tier)

Tier	Method	Speed	Hallucination Risk
1. Template Lookup	18+ pre-written safe alternatives based on FTC/EU/CFPB standards	Instant	Zero
2. Keyword Matching	Match clause text to relevant templates via legal keywords	Instant	Zero
3. LLM Refinement	Qwen2.5-7B adapts template to specific clause context	~3-5s	Low (template-anchored)

Anti-hallucination guardrails:

• Template anchor: LLM can only refine, not generate from scratch
• Legal citation: Every suggestion includes legal basis and consumer standard
• Disclaimer: Clear "Not legal advice" warning

📊 Risk Scoring Methodology

Risk scores combine clause detection with weighted severity:

• CRITICAL: 40 pts (Uncapped Liability, Arbitration, IP Assignment, etc.)
• HIGH: 20 pts (Non-Compete, Exclusivity, Unilateral Change, etc.)
• MEDIUM: 10 pts (Governing Law, Jurisdiction, etc.)
• LOW: 3 pts (Document Name, Dates, etc.)

Final score normalized to 0-100 with letter grades:

• A (0-14): Low risk
• B (15-29): Moderate risk
• C (30-49): Elevated risk
• D (50-69): High risk
• F (70+): Critical risk

🚀 Usage

1. Upload a contract (PDF, DOCX, or TXT) or paste text directly
   • 💡 Scanned PDFs are automatically processed with OCR
2. Click Analyze Contract
3. View results across tabs:
   • Document: Full text with inline entity highlights
   • Clauses: Detected clauses with risk badges
   • Entities: Extracted parties, dates, money, jurisdictions
   • Contradictions: Conflicting clauses and missing provisions
   • Obligations: Action items categorized by type
   • Compliance: Regulatory framework checks
   • Redlining: ✏️ Safer clause alternatives with legal citations
4. Export JSON/CSV reports
5. Switch to 💬 Contract Q&A tab to ask questions about your contract

🔀 Compare Contracts

Switch to the Compare Contracts tab to:

• Upload or paste two contracts side-by-side
• See clause-level diffs (added, removed, modified)
• Get an alignment score and risk delta

⚠️ Disclaimer

Not legal advice. ClauseGuard is an AI-powered analysis tool for informational purposes only. Always consult a qualified attorney for legal decisions. The tool may miss nuances and should be used as a preliminary screening aid, not a substitute for professional legal review.

🔗 Links

• ClauseGuard Space
• Clause Classifier Model
• Legal-BERT Base
• CUAD Dataset
• Qwen2.5-7B (Chatbot LLM)
• docTR OCR
• CUAD Paper (arXiv:2103.06268)

---

Built with ❤️ using Gradio, Hugging Face Transformers, and Legal-BERT. Open source and free for all.