🛡️ Kavach (कवच)

The Enterprise Privacy Shield for Large Language Models

Kavach is an open-source, high-performance privacy middleware designed to sit between your sensitive data and public LLMs (like OpenAI, Claude, or Gemini). It ensures that Personally Identifiable Information (PII) never leaves your secure infrastructure.

🏗️ How It Works

Kavach acts as a secure "middleman". It intercepts your prompt, uses a local Small Language Model (SLM) to identify sensitive data, replaces it with tokens, helps you get the answer from the LLM, and then restores the original data in the response.

Workflow Diagram

sequenceDiagram
    participant User as 👤 User/Client
    participant Kavach as 🛡️ Kavach Core
    participant Redis as 💾 Redis (State)
    participant LLM as 🤖 Public LLM

    %% Masking Flow
    User->>Kavach: "Is Amit eligible for 50L loan?"
    note right of Kavach: 1. Identify PII (Amit -> PERSON)
    Kavach->>Redis: Store Mapping { "{{PERSON_1}}": "Amit" }
    Kavach->>Kavach: Mask: "Is {{PERSON_1}} eligible for 50L loan?"
    Kavach->>LLM: Send Masked Prompt

    %% Unmasking Flow
    LLM-->>Kavach: "{{PERSON_1}} is eligible."
    note right of Kavach: 2. Lookup Token ({{PERSON_1}})
    Kavach->>Redis: Retrieve "Amit"
    Kavach->>User: "Amit is eligible."

🚀 Getting Started

You can use Kavach in two ways: as a Node.js Library or as a Standalone Microservice.

Option A: Use as a Library (Node.js/TypeScript)

Best for integrating directly into your existing backend.

1. Install:

   npm install @karanbavari/kavach

2. Usage:

   import { Kavach } from '@karanbavari/kavach';
   
   // Initialize (Using Memory Default, Redis recommended for prod)
   const kavach = new Kavach();
   const sessionId = 'user-session-123';
   
   // 1. Sanitize Input
   const sensitiveInput = "Please diagnose patient John Doe, born 1980-01-01.";
   const safePrompt = await kavach.sanitize(sensitiveInput, sessionId);
       
   console.log(safePrompt); 
   // Output: "Please diagnose patient {{PERSON_1}}, born {{DATE_1}}."
   
   // ... [SEND safePrompt TO OPENAI/LLM] ...
   const llmResponse = "The patient {{PERSON_1}} shows symptoms of...";
   
   // 2. Desanitize Output
   const finalResponse = await kavach.desanitize(llmResponse, sessionId);
       
   console.log(finalResponse);
   // Output: "The patient John Doe shows symptoms of..."

Option B: Use as a Microservice (Docker)

Best for Microservices, Python, Go, or Java environments.

1. Run with Docker Compose:

   docker-compose up --build

2. API Usage:

   Request (Sanitize):

   curl -X POST http://localhost:3000/v1/sanitize \
     -H "Content-Type: application/json" \
     -d '{"text": "Call Sarah at Google", "sessionId": "sess-1"}'

   Response:

   { "masked_text": "Call {{PERSON_1}} at {{ORG_1}}" }

🌍 Top 20 Use Cases

Kavach is designed to be industry-agnostic. Here is how it can be applied across top 20 software domains:

| Category | Example Application | Use Case for Kavach | | :--- | :--- | :--- | | 1. Healthcare (MedTech) | Patient Diagnosis Bots | Removing Patient Names (PHI) before asking GPT-4 for diagnosis analysis. | | 2. FinTech | Loan Approval Systems | Masking User Income, Name, and Credit Score before risk assessment. | | 3. LegalTech | Contract Review AI | Hiding Client Names and specific monetary values in legal contracts. | | 4. HR & Recruitment | Resume Analyzers | Removing Candidate Name/Email/Phone to reduce bias and protect privacy. | | 5. E-commerce | Customer Support Chatbots | Masking Order IDs and User Addresses during support queries. | | 6. EdTech | Personalized Tutors | Protecting Student Identity and Grade records. | | 7. CRM Systems | Lead Scoring | Anonymizing contact details of sales leads while analyzing sentiment. | | 8. Cyber Security | Incident Reporting | Masking internal IP addresses and server names in log analysis. | | 9. Insurance | Claim Processing | Hiding Policy Numbers and Claimant details. | | 10. Real Estate | Property Valuation | Masking current Owner Name and exact address coordinates. | | 11. Mental Health | Therapy Chatbots | Extreme privacy for user confessions and names. | | 12. Travel & Tech | Booking Assistants | Protecting Passport Numbers and Flight PNRs. | | 13. Supply Chain | Vendor Negotiator Bots | Hiding Vendor Names and specific quote prices during analysis. | | 14. Social Media | Content Moderation | Anonymizing User Handles in flagged content review. | | 15. Public Sector | Citizen Services | Protecting National ID numbers (like Aadhaar/SSN). | | 16. Journalism | Source Protection | Anonymizing identities of whistleblowers in transcript analysis. | | 17. Automotive | Connected Car Assistants | Masking Driver Location and Home Address history. | | 18. Telecommunications | Call Center AI | Redacting phone numbers and account pins from call transcripts. | | 19. Marketing Tech | User Segmentation | Analyzing behavior without exposing PII of target lists. | | 20. Research/Academia | Survey Analysis | Ensuring participant anonymity in large-scale qualitative surveys. |

📜 Open Source License

Kavach is proudly Open Source and licensed under the MIT License.

What this means for you:

• ✅ Free to Use: You can use Kavach in your personal, educational, or commercial projects for free.
• ✅ Modification: You can modify the source code to fit your specific needs.
• ✅ Distribution: You can distribute your modified versions.
• ✅ No Warranty: The software is provided "as is".

We believe privacy is a fundamental right, and building secure AI tools should be accessible to everyone.

🤝 Contributing to Kavach

We welcome contributions from the community! Whether it's fixing bugs, adding new PII detectors, or improving documentation.

How to Contribute

1. Fork the Repository: Click the "Fork" button on the top right of this page.
2. Clone your Fork:

   git clone https://github.com/YOUR_USERNAME/kavach.git
   cd kavach

3. Create a Branch:

   git checkout -b feature/amazing-new-feature

4. Make Changes: Write your code and add tests.
5. Commit & Push:

   git commit -m "Add amazing new feature"
   git push origin feature/amazing-new-feature

6. Open a Pull Request: Go to the original Kavach repository and click "Compare & pull request".

See CONTRIBUTING.md for detailed guidelines.

👨‍💻 Created By

Karan J Solanki
Chief Technical Officer (CTO)
BotPey Bharat AI Innovation & Solutions Pvt. Ltd.

Built with ❤️ for the Open Source Community.