Persistent-Code MCP Server

# LlamaIndex Integration Implementation This document describes the implementation of the LlamaIndex integration for the Persistent-Code MCP server. ## Overview The LlamaIndex integration enhances the Persistent-Code MCP server with semantic search capabilities, allowing Claude to understand code based on meaning rather than just keywords. ## Implementation The implementation consists of the following components: 1. **LlamaIndexManager**: A new module that encapsulates all LlamaIndex functionality 2. **KnowledgeGraph**: Updated to use the LlamaIndexManager for semantic operations 3. **Test Scripts**: Added examples to demonstrate the semantic search capabilities 4. **Documentation**: Updated to explain the LlamaIndex integration ## Benefits The LlamaIndex integration provides the following benefits: 1. **Semantic Search**: Find code components based on their meaning 2. **Vector Embeddings**: Efficient similarity matching for code components 3. **Knowledge Graph**: Graph-based representation of code relationships 4. **Contextual Retrieval**: More relevant code suggestions ## Usage To use the semantic search capabilities: 1. Ensure LlamaIndex is enabled in the config ```python from persistent_code.config import config config.set("llama_index", "enabled", True) ``` 2. Use the search_code method with natural language queries ```python from persistent_code.knowledge_graph import KnowledgeGraph kg = KnowledgeGraph("my_project") results = kg.search_code("function to validate user input") ``` 3. Try the semantic search CLI ```bash python semantic_search_cli.py --project my_project --semantic --query "function to validate user input" ``` ## Architecture The architecture follows these design principles: 1. **Separation of Concerns**: LlamaIndex functionality is isolated in its own manager 2. **Error Handling**: Robust error handling for graceful fallback to basic search 3. **Configuration**: Easy configuration of LlamaIndex features 4. **Backward Compatibility**: Maintains compatibility with existing code ## Testing To test the LlamaIndex integration: ```bash python examples/test_semantic_search.py --project test_project --file examples/sample_code.py --query "function to handle authentication" ``` ## Future Improvements Potential future improvements include: 1. More sophisticated semantic triple extraction 2. Support for different embedding models 3. Caching of embeddings for better performance 4. Integration with external embedding services