Table of Contents

👉 Try the demo: https://github.com/LM-Kit/lm-kit-net-samples/tree/main/console_net/contract_key_terms_extractor

AI Contract Key Terms Extractor for C# .NET Applications


🎯 Purpose of the Demo

AI Contract Key Terms Extractor demonstrates how to use LM-Kit.NET with vision-language models (VLMs) to extract key clauses and terms from contracts entirely on local hardware. Feed it any legal agreement, whether plain text, PDF, DOCX, or a scanned contract image, and get back structured JSON with parties, dates, financials, termination conditions, IP rights, confidentiality obligations, liability caps, and risk flags.

The sample shows how to:

  • Use TextExtraction with VLMs to extract structured data from legal documents and scanned contract images.
  • Define extraction schemas with TextExtractionElement for contract-specific fields.
  • Extract both factual data (dates, amounts) and interpretive analysis (risk flags, clause summaries).
  • Process text content with SetContent(string) or load documents and images with Attachment.

Why a Local Contract Analyzer with LM-Kit.NET?

  • Scanned contract support: VLMs process photographed or scanned contracts directly, no separate OCR step required.
  • Confidential contracts stay private: analyze sensitive legal documents without cloud exposure.
  • One API call: a single Parse() call extracts all key terms. No agents, no chaining.
  • Risk detection: AI automatically flags unusual or potentially risky clauses.
  • Works offline: no API keys, no internet connection required.

👥 Who Should Use This Demo

  • Legal Teams: quickly review contracts and identify key terms without manual reading, even from scanned documents.
  • Procurement Departments: extract and compare terms across vendor agreements, including photographed contracts.
  • Startup Founders: understand key clauses before signing partnerships or vendor contracts.
  • Contract Management Developers: integrate local AI-powered contract analysis into CLM systems.
  • Compliance Officers: flag risky clauses and ensure policy adherence.

🚀 What Problem It Solves

  • Contract review time: extract key terms in seconds instead of hours of manual reading.
  • Scanned documents: process photographed or scanned contracts directly with vision-language models.
  • Data privacy: analyze confidential contracts without uploading to cloud AI services.
  • Risk visibility: automatically surface non-compete, exclusivity, auto-renewal, and penalty clauses.
  • Consistency: extract the same structured fields from every contract for easy comparison.

💻 Demo Application Overview

Console app that:

  • Lets you choose from multiple vision-language models optimized for document understanding.
  • Downloads models if needed, with live progress updates.
  • Creates a TextExtraction instance with a contract-specific schema.
  • Enters an interactive loop where you can:
    • Type sample to analyze a built-in Master Services Agreement.
    • Enter a file path to analyze your own contract (.txt, .pdf, .docx, .png, .jpg).
    • View the extracted key terms with color-coded output.
    • View the full JSON output for integration.
  • Loops until you type q.

Key Features

  • Vision-Language Models: Process scanned contracts, photographs, and standard text formats using VLMs.
  • Comprehensive Extraction: Contract type, parties, dates, value, payment, termination, IP, confidentiality, liability, governing law.
  • Risk Flags: AI identifies unusual or risky clauses (non-compete, exclusivity, unlimited liability, auto-renewal).
  • Party Identification: Each party extracted with name, role, and address.
  • Built-In Sample: Realistic Master Services Agreement for instant demo.
  • JSON Output: Machine-readable structured data for contract management systems.
  • Zero Configuration: No API keys, no cloud accounts, no external dependencies.

🏗️ Architecture

  ┌─────────────────────────────────────────────────┐
  │   Contract (text, PDF, DOCX, scanned image)     │
  └───────────────────┬─────────────────────────────┘
                      │
                      ▼
  ┌──────────────────────────────────────────────────┐
  │         Vision-Language Model (VLM)              │
  └───────────────────┬──────────────────────────────┘
                      │
                      ▼
  ┌──────────────────────────────────────────────────┐
  │              TextExtraction                      │
  │                                                  │
  │   Schema:                                        │
  │   ├── Contract Type (String)                     │
  │   ├── Contract Title (String)                    │
  │   ├── Parties (Array)                            │
  │   │   └── Name, Role, Address                    │
  │   ├── Effective Date (Date)                      │
  │   ├── Expiration Date (Date)                     │
  │   ├── Contract Value (String)                    │
  │   ├── Payment Terms (String)                     │
  │   ├── Termination Clause (String)                │
  │   ├── Confidentiality (String)                   │
  │   ├── Liability Limitation (String)              │
  │   ├── Intellectual Property (String)             │
  │   ├── Governing Law (String)                     │
  │   ├── Key Obligations (String)                   │
  │   ├── Renewal Terms (String)                     │
  │   └── Risk Flags (String)                        │
  │                                                  │
  └───────────────────┬──────────────────────────────┘
                      │
                      ▼
  ┌──────────────────────────────────────────────────┐
  │           Key Contract Terms                     │
  │           (formatted display + JSON)             │
  └──────────────────────────────────────────────────┘

Extraction Schema

Field Type Description
Contract Type String NDA, Service Agreement, Employment, Lease, etc.
Contract Title String Official document title
Parties Array Name, role, and address of each party
Effective Date Date When the contract takes effect
Expiration Date Date End date or "Perpetual"
Contract Value String Total monetary value or fee
Payment Terms String Payment schedule and conditions
Termination Clause String Termination conditions and notice period
Confidentiality String Non-disclosure obligations summary
Liability Limitation String Liability cap or damages limitation
Intellectual Property String IP ownership and licensing provisions
Governing Law String Jurisdiction for disputes
Key Obligations String Main deliverables and obligations
Renewal Terms String Auto-renewal conditions
Risk Flags String Unusual or risky clauses identified

Code Highlights

using LMKit.Extraction;
using LMKit.Model;

LM model = LM.LoadFromModelID("qwen3.5:9b");

var textExtraction = new TextExtraction(model);
textExtraction.Elements = new List<TextExtractionElement>
{
    new("Contract Type", ElementType.String, "Type of contract."),
    new("Parties",
        new List<TextExtractionElement>
        {
            new("Name", ElementType.String, "Legal name of the party."),
            new("Role", ElementType.String, "Role in the contract."),
            new("Address", ElementType.String, "Registered address.")
        },
        isArray: true,
        "Parties involved in the contract."
    ),
    new("Effective Date", ElementType.Date, "When the contract becomes effective."),
    new("Risk Flags", ElementType.String, "Unusual or risky clauses."),
    // ... more elements
};

// Text content
textExtraction.SetContent(contractText);

// Or scanned contract image / PDF
textExtraction.SetContent(new Attachment("contract_scan.png"));

var result = textExtraction.Parse();
Console.WriteLine(result.Json);

⚙️ Getting Started

Prerequisites

  • .NET 8.0 or later
  • Sufficient VRAM for the selected model (2.5-18 GB depending on model choice)

Download

git clone https://github.com/LM-Kit/lm-kit-net-samples
cd lm-kit-net-samples/console_net/contract_key_terms_extractor

Run

dotnet build
dotnet run

Then:

  1. Select a vision-language model from the menu, or paste a custom model URI.
  2. Wait for the model to download (first run) and load.
  3. Type sample to try the built-in contract, or enter a file path (text, PDF, DOCX, or scanned image).
  4. View the extracted key terms and JSON output.
  5. Type q to exit.

🔧 Troubleshooting

  • Incomplete clause extraction

    • Try a larger model (12B+) for better accuracy on complex legal language.
    • Ensure the full contract text is provided, not just excerpts.

  • Risk flags not detected

    • Larger models are better at identifying subtle risk patterns.
    • Check that the contract contains the relevant clauses.

  • Scanned contract issues

    • Ensure the image is clear and well-lit for best VLM results.
    • Higher resolution images produce better extraction accuracy.
    • Try a larger VLM if extraction from images is incomplete.

🚀 Extend the Demo

  • Contract comparison: extract terms from multiple contracts and generate a comparison table.
  • Compliance checking: validate extracted terms against company policy rules.
  • Renewal tracking: build a dashboard of upcoming contract expirations and renewals.
  • Multi-language contracts: test with international agreements using multilingual VLMs.
  • Database integration: pipe JSON output into a contract lifecycle management (CLM) system.
  • Batch processing: scan a folder of contract images and extract terms from each.

📚 Additional Resources

📚 Related Content

Share