🧠 Archon Knowledge: Your AI's Memory Bank

Build a powerful knowledge base for your AI assistants. Crawl websites, upload documents, and give your AI instant access to all your technical and business information.

Archon Knowledge transforms your documentation, websites, and files into a searchable knowledge base that your AI coding assistants can instantly access. Never explain the same concept twice - your AI remembers everything.

🎉Fully Operational RAG System

The RAG system is now fully functional with 14 MCP tools enabled, comprehensive error handling, and threading optimizations for high performance.

🏗️ How RAG Works

⚡ Performance Features

Archon Knowledge is optimized for speed and efficiency:

• Smart Concurrency: Adaptive processing based on system resources
• Batch Processing: Processes multiple documents efficiently
• Rate Limiting: Respects API limits while maximizing throughput
• Memory Management: Automatically adjusts to available system memory

🔍 Using the Knowledge Base

Basic Search

The perform_rag_query tool is the primary interface for semantic search across your knowledge base:

Basic RAG Query

// Simple search across all sources
await mcp.callTool('perform_rag_query', {
  query: "authentication best practices",
  match_count: 5  // Optional, defaults to 5
});

Filtered Search by Source

Filter results to specific domains or sources:

Source-Filtered Search

// Search only within a specific domain
await mcp.callTool('perform_rag_query', {
  query: "MCP session management",
  source: "modelcontextprotocol.io",  // Filter by domain
  match_count: 10
});

// Get available sources first
const sources = await mcp.callTool('get_available_sources', {});
// Returns: ["ai.pydantic.dev", "modelcontextprotocol.io", ...]

Advanced Usage Examples

Technical Documentation

// Search for technical implementation details
await mcp.callTool('perform_rag_query', {
  query: "SSE transport implementation MCP protocol",
  source: "modelcontextprotocol.io",
  match_count: 5
});

// Response includes:
// - Matched content chunks
// - Source URLs
// - Similarity scores
// - Metadata (headers, context)

Code Examples

// Search for code examples
await mcp.callTool('search_code_examples', {
  query: "React hooks useState useEffect",
  source_id: "react.dev",  // Optional source filter
  match_count: 10
});

// Returns:
// - Code snippets with syntax highlighting
// - AI-generated summaries
// - Full context (before/after code)
// - Source file information

Multi-Source Search

// Search across all indexed sources
const results = await mcp.callTool('perform_rag_query', {
  query: "best practices for API design REST GraphQL",
  // No source filter - searches everything
  match_count: 15
});

// Group results by source
const groupedResults = results.reduce((acc, result) => {
  const source = result.metadata.source;
  if (!acc[source]) acc[source] = [];
  acc[source].push(result);
  return acc;
}, {});

🔧 Advanced Features

• Contextual Embeddings: Enhanced understanding through document context
• Source Filtering: Search within specific domains or documentation sources
• Code Search: Specialized search for code examples and implementations
• Multi-Source: Search across all your indexed knowledge sources simultaneously

⚡ Performance

📊Fast & Efficient

• Average Query Time: 200-300ms
• Optimized Processing: Smart batching and concurrency
• Memory Adaptive: Automatically adjusts to system resources
• Rate Limited: Respects API limits for reliable operation

📊 Real-Time Progress

When processing large amounts of content, Archon provides real-time progress updates via Socket.IO:

• Smooth Progress: Linear progression from 0-100%
• Batch Details: Clear information about processing status
• Real-Time Updates: Live updates as documents are processed
• Memory Awareness: Automatically adjusts based on system resources

🗄️ Data Storage

Archon uses a vector database to store and search your knowledge:

• Vector Embeddings: Content is converted to high-dimensional vectors for semantic search
• Source Tracking: Each document is linked to its original source
• Code Examples: Special handling for code snippets with language detection
• Metadata Storage: Additional context and headers are preserved

🔧 Common Issues

Performance

• Slow searches: Usually due to large document sets - the system automatically optimizes batch sizes
• Memory usage: Adaptive processing automatically adjusts based on available system memory
• Rate limiting: Built-in rate limiting prevents API quota issues

Search Quality

• Poor results: Try different search terms or use source filtering to narrow results
• Missing content: Ensure documents are properly crawled and indexed
• Code examples: Use the specialized search_code_examples tool for better code results

🚀 Getting Started

1. Add Knowledge Sources: Use MCP tools to crawl websites and upload documents
2. Search Your Knowledge: Use perform_rag_query to find relevant information
3. Filter by Source: Search within specific domains when you need focused results
4. Find Code Examples: Use search_code_examples for code-specific searches

🔮 What's Next

Future enhancements include multi-model processing, hybrid search combining vector and keyword search, and advanced neural reranking for even better results.