Innovaas KMS Server

Enhanced Model Context Protocol (MCP) server for the Innovaas RAG Knowledge Management System. This server exposes powerful multi-modal search, RAG-powered chat with intelligent token management, and comprehensive document access to external systems via the standardized MCP protocol.

⚡ Latest v1.0.0 Features

🎯 Intelligent Token Management

• Automatic Optimization: Prevents API token limit errors (65K+ → 30K tokens)
• Provider-Aware: Different limits for OpenAI (30K) vs Claude (200K)
• Smart Document Selection: Prioritizes by relevance, includes summaries of excluded content
• Zero Configuration: Works automatically with kms_chat tool

🔍 Advanced Search Capabilities

• Full Document Content: Complete text (4,000+ characters) instead of 200-char previews
• Multi-Modal Search: Text, audio transcriptions, video frames, and technical content
• Intelligent Routing: Enhanced RAG with query analysis and optimal strategy selection
• Technical Content Detection: Find code, diagrams, and UI elements in video content

💬 Enhanced RAG-Powered Chat

• Comprehensive Responses: Based on complete source material with full content access
• Source Citations: Precise document and timestamp references
• Provider Choice: OpenAI GPT-4o-mini or Claude for different use cases
• Context Filtering: Focus conversations by tags and document types

🚀 Quick Start

1. Installation

# Clone the repository
git clone https://github.com/innovaas/kms-mcp-server.git
cd kms-mcp-server

# Install dependencies
npm install

# Build the server
npm run build

2. Configuration

# Required: KMS API endpoint
export KMS_BASE_URL="https://your-kms-domain.com/kms"

# Required: Authentication key
export BACKGROUND_PROCESS_API_KEY="your-secure-api-key"
# OR use MCP-specific key
export MCP_API_KEY="your-mcp-api-key"

3. Run the Server

# Development mode
npm run dev

# Production mode
npm start

# With environment variables inline
KMS_BASE_URL="https://your-domain.com/kms" BACKGROUND_PROCESS_API_KEY="your-key" npm start

🛠️ Integration Examples

Claude Desktop Configuration

Add to your Claude Desktop config file (~/.claude_desktop_config.json):

{
  "mcpServers": {
    "innovaas-kms": {
      "command": "node",
      "args": ["/path/to/kms-mcp-server/dist/index.js"],
      "env": {
        "KMS_BASE_URL": "https://your-domain.com/kms",
        "BACKGROUND_PROCESS_API_KEY": "your-secure-api-key"
      }
    }
  }
}

Cline/VSCode Integration

Configure in your MCP settings:

{
  "name": "innovaas-kms",
  "serverPath": "/path/to/kms-mcp-server/dist/index.js",
  "environment": {
    "KMS_BASE_URL": "https://your-domain.com/kms",
    "MCP_API_KEY": "your-secure-api-key"
  }
}

Programmatic Integration

import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";

const transport = new StdioClientTransport({
  command: "node",
  args: ["/path/to/kms-mcp-server/dist/index.js"],
  env: {
    KMS_BASE_URL: "https://your-domain.com/kms",
    MCP_API_KEY: "your-api-key"
  }
});

const client = new Client(
  { name: "kms-client", version: "1.0.0" },
  { capabilities: {} }
);

await client.connect(transport);

// Use intelligent search with full content
const result = await client.callTool({
  name: "kms_intelligent_search",
  arguments: {
    query: "What are the best practices for implementing a Unified Namespace?",
    maxResults: 10
  }
});

🎯 Available Tools

kms_chat 🚀 Primary Tool

Comprehensive knowledge queries with intelligent token management

{
  "message": "How do I implement OEE monitoring in a manufacturing environment?",
  "provider": "openai",
  "useMultiModal": true,
  "tags": ["OEE", "manufacturing"],
  "maxResults": 15
}

✅ Key Benefits:

• Token Optimization: Automatically prevents API limit errors
• Full Content Access: Complete document text (4,000+ characters)
• Provider-Aware: Adjusts context size for OpenAI vs Claude
• Multi-Modal Context: Combines text, video, and web sources

kms_intelligent_search

Advanced RAG search with query analysis

{
  "query": "unified namespace MQTT implementation patterns",
  "maxResults": 15,
  "filters": {
    "type": "video",
    "tags": ["UNS", "MQTT"]
  },
  "includeAnalysis": true
}

kms_multimodal_search

Search across all content types

{
  "query": "user authentication flow diagrams",
  "searchMode": "multimodal",
  "maxResults": 10,
  "filters": {
    "hasVisualContent": true,
    "documentTypes": ["video", "whitepaper"]
  }
}

kms_search

Basic semantic search

{
  "query": "manufacturing execution systems",
  "limit": 10,
  "threshold": 0.7
}

kms_get_document

Retrieve specific document

{
  "documentId": "uuid-of-document"
}

kms_get_stats

System analytics

{
  "includeProcessingDetails": true
}

kms_list_documents

Browse documents

{
  "limit": 25,
  "type": "video",
  "tags": ["training", "technical"],
  "mediaType": "video"
}

🎉 What's Fixed in v1.0.0

❌ Before: Token Limit Errors

Error: Request too large for gpt-4o: Limit 30000, Requested 70239

✅ After: Intelligent Optimization

{
  "tokenOptimization": {
    "enabled": true,
    "documentsIncluded": 8,
    "documentsExcluded": 7,
    "optimization": "Included 8/15 documents, using ~27,518 tokens",
    "estimatedTotalTokens": 27518
  }
}

🔧 Improvements Made

1. Automatic Token Management: No more API limit errors
2. Smart Document Selection: Prioritizes most relevant content
3. Full Content Access: 4,000+ character responses vs 200-char previews
4. Provider Optimization: Different strategies for OpenAI vs Claude
5. Transparent Operation: Shows what was included/excluded and why

📊 System Capabilities

Current KMS Status ✅

• 127+ documents processed with 100% success rate
• 1,000+ video frames extracted and analyzed
• Multi-modal search across text, audio, and video
• Technical content detection for code, diagrams, UI elements
• Real-time processing pipeline with error recovery

Content Coverage

• Technical Documentation: API docs, system architecture, code examples
• Training Videos: 105+ processed videos with transcription and frame analysis
• Manufacturing Content: MES, OEE, UNS, MQTT, IoT, SCADA terminology
• Web Resources: Crawled documentation and technical resources

AI Capabilities

• AssemblyAI: High-quality transcription with technical term boosting
• OpenAI Embeddings: 1536-dimensional vectors for semantic search
• Claude Vision: Technical content analysis for diagrams and code
• Multi-Provider Chat: OpenAI GPT-4o-mini and Claude support

🛡️ Authentication & Security

API Key Authentication

# Set authentication key
export BACKGROUND_PROCESS_API_KEY="secure-random-string"

# Or use MCP-specific key
export MCP_API_KEY="mcp-specific-secure-key"

Network Configuration

• Protocol: HTTPS (secure connection)
• Transport: STDIO (standard for MCP)
• Authentication: Bearer token with API key

📋 Development

Project Structure

kms-mcp-server/
├── src/
│   └── index.ts           # Main MCP server implementation
├── dist/                  # Built files (generated by npm run build)
├── examples/              # Configuration examples
├── package.json           # Dependencies and scripts
├── tsconfig.json          # TypeScript configuration
└── README.md             # This file

Scripts

npm run build              # Compile TypeScript to JavaScript
npm run dev                # Development mode with hot reload
npm start                  # Run compiled server
npm run clean              # Clean build directory
npm test                   # Run tests

Requirements

• Node.js: 18.0.0 or higher
• TypeScript: 5.0.0 or higher
• KMS Server: Running Innovaas KMS instance

🐛 Troubleshooting

Common Issues

1. Connection Failed

   Error: KMS API request failed: 500 Internal Server Error
   

   • ✅ Ensure KMS server is running
   • ✅ Check KMS_BASE_URL environment variable
   • ✅ Verify network connectivity

2. Authentication Errors

   Error: 401 Unauthorized
   

   • ✅ Verify API key is set correctly
   • ✅ Check Bearer token format
   • ✅ Ensure KMS server has matching API key

3. Token Limit Errors (Should be fixed)

   Error: Request too large for gpt-4o: Limit 30000, Requested 65879
   

   • ✅ Update to v1.0.0 with token optimization
   • ✅ Use kms_chat tool (automatically optimized)
   • ✅ Check tokenOptimization in responses

Debug Mode

# Enable verbose logging
DEBUG=1 npm run dev

# Check KMS server status
curl -H "Authorization: Bearer your-api-key" https://your-domain.com/kms/api/dashboard-stats

🤝 Contributing

1. Fork the repository
2. Create a feature branch: git checkout -b feature/amazing-feature
3. Make your changes
4. Run tests: npm test
5. Build: npm run build
6. Commit changes: git commit -m 'Add amazing feature'
7. Push to branch: git push origin feature/amazing-feature
8. Create Pull Request

Development Guidelines

• Follow existing code patterns for consistency
• Add comprehensive error handling
• Update tool schemas when modifying parameters
• Test with multiple MCP clients before committing
• Document new features in README

📄 License

MIT License - see the LICENSE file for details.

🔗 Links

• GitHub Repository: https://github.com/innovaas/kms-mcp-server
• Issues: https://github.com/innovaas/kms-mcp-server/issues
• Innovaas Website: https://innovaas.co
• Model Context Protocol: https://modelcontextprotocol.io

---

🚀 Ready to integrate your knowledge management with any MCP-compatible system with intelligent token optimization!