v0.4.0

Minor Changes

• b4a6844: Enhanced Codebase Indexing and Search Capabilities

  This release brings significant improvements to code parsing, indexing quality, and search performance.

  Intelligent Indexing Improvements

  • Chunk Quality Filtering: Automatically filters out 20-40% of low-quality code chunks (closing tags, empty sections, boilerplate) before indexing, improving search relevance
  • Embedding Enrichment: Adds contextual metadata (file path, type, name, parameters) to code embeddings for better semantic understanding
  • Advanced Path Optimization: Implements path segment decomposition for 6.7x faster directory-scoped searches
  • HNSW Tuning: Optimized vector search parameters (m=64, ef_construct=512, hnsw_ef=128) for 30-40% better recall

  Enhanced Language Support

  • Java Extractor: Full support for classes, interfaces, enums, methods, constructors, and fields with modifiers, generics, and exceptions
  • Go Extractor: Comprehensive extraction of functions, methods, structs, interfaces, type aliases, and exported symbols
  • Kotlin & Swift: Added language registry support for .kt, .kts, and .swift files

  TypeScript/JavaScript Parser Enhancements

  • Test Framework Detection: Extracts test blocks (describe, it, test, beforeEach, afterEach, etc.) with descriptions for better test suite indexing
  • Namespace/Module Extraction: Proper extraction of TypeScript namespace and module declarations
  • Interface Method Signatures: Extracts individual methods from interfaces for granular API documentation
  • Public Field Extraction: Captures class properties with visibility, static, and readonly modifiers
  • JSX Filtering: Intelligently filters pure JSX markup while preserving components with event handlers, hooks, or conditional logic
  • Enhanced Method Detection: Identifies constructors, getters, setters, and abstract methods with rich metadata
  • Abstract Class Support: Properly detects and marks abstract class declarations

  Search Performance

  • Directory Filtering: New directoryPrefix parameter for limiting searches to specific directories (e.g., only search in "src/components")
  • Payload Indexes: Created efficient indexes on type, language, and path segments for faster filtering
  • Metadata Exclusion: Automatically excludes metadata points from search results

  Bug Fixes

  • Fixed unnecessary escape characters in regex patterns for linter compliance

  Breaking Changes

  None - all changes are backward compatible.

v0.3.0

Minor Changes

• de79027: Fix indexing status and add major performance optimizations (5-10x faster)

  Bug Fixes:

  • Fixed indexing status MCP command returning all zeros for statistics (totalFiles, totalBlocks, totalVectors)
  • Status now correctly updates from storage regardless of system state

  Performance Improvements (5-10x faster indexing):

  • NEW: Cross-file embedding batching - Batch embeddings across multiple files instead of per-file (5-10x faster)
  • NEW: Cross-file storage batching - Batch Qdrant upserts across files (2-3x faster)
  • NEW: Parallel file discovery - Concurrent directory traversal (2-4x faster)
  • NEW: OptimizedBatchProcessor - New processor with all optimizations enabled by default
  • Fixed critical memory leak in batch processor concurrency control (60% memory reduction)
  • Optimized file change detection to use stats instead of reading entire file contents
  • Added explicit memory cleanup between processing cycles
  • Reduced default batch sizes (concurrency: 5→3, batch size: 50→20)
  • Added pauses between batches to allow garbage collection

  New Features:

  • useOptimizations config option (defaults to true)
  • Cross-file batching with configurable batch sizes
  • Parallel file discovery with configurable concurrency
  • Comprehensive performance documentation in PERFORMANCE.md

  Breaking Changes: None - optimizations are enabled by default but fully backwards compatible

  These changes provide 5-10x faster indexing with 60% less memory usage while maintaining stability.

v0.2.2

Patch Changes

• 6f6db4b: Fix README package name from @itseasy21/mcp-codebase-index to mcp-codebase-index to match npm registry

v0.2.1

Patch Changes

• 29326c0: Transform README into professional MCP package documentation with comprehensive IDE/editor setup instructions. Added configuration guides for Claude Desktop, Claude Code, Cursor, Continue, Windsurf, Zed, and other MCP-compatible clients. Removed development-specific content and focused on user experience with clear installation steps, usage examples, and troubleshooting guides.

v0.2.0

Minor Changes

• 806223b: Complete MCP Codebase Index Implementation (Phases 1-10)

  This is the initial complete implementation of the MCP Codebase Index server, providing semantic code search capabilities through the Model Context Protocol.

  🎉 Phase 1: Foundation

  • TypeScript project setup with strict mode and ESM modules
  • Zod-based configuration system with environment variable support
  • Comprehensive utility functions (logger, error classes, retry logic, file operations)
  • Basic MCP server infrastructure with placeholder handlers

  📝 Phase 2: Code Parsing

  • Tree-sitter integration for TypeScript, Python, JavaScript, Go, Rust, C/C++, C#, Java, Ruby, PHP
  • Markdown parser with header-based chunking
  • Fallback line-based chunking for unsupported languages
  • Comprehensive code block extraction with metadata

  🧠 Phase 3: Embedding Generation

  • Multi-provider embedding system supporting:
    • Google Gemini (text-embedding-004)
    • OpenAI (text-embedding-3-small/large)
    • Ollama (local models)
    • OpenAI-compatible services
  • LRU caching for embedding results
  • Batch processing with retry logic
  • Provider health checks and validation

  💾 Phase 4: Vector Storage

  • Qdrant vector database integration
  • Collection management with automatic creation
  • Batch upsert, delete, and scroll operations
  • Semantic search with filtering and pagination
  • Point metadata storage and retrieval

  🔄 Phase 5: Indexing Engine

  • Complete end-to-end indexing pipeline
  • File watching with debouncing
  • Git branch change detection
  • Priority-based task queue with retries
  • Concurrent batch processing
  • File hash caching for incremental updates
  • .mcpignore and .gitignore support

  🔍 Phase 6: Search Implementation

  • Semantic search with vector similarity
  • Multi-factor relevance ranking:
    • Exact match detection
    • Name similarity scoring
    • File path relevance
    • Recency scoring
    • Language and type boosting
  • Context extraction with configurable line ranges
  • Result caching with TTL
  • Filter support (file types, paths, languages)

  📊 Phase 7: Status Management

  • State machine for indexing status (standby, indexing, indexed, error)
  • Real-time progress tracking
  • Statistics collection (blocks, vectors, languages)
  • Error aggregation and reporting
  • Progress percentage calculation

  🔧 Phase 8: MCP Tools Integration

  • Orchestrator class wiring all components together
  • Six MCP tool handlers:
    1. codebase_search - Semantic code search with filters
    2. indexing_status - Real-time status and statistics
    3. reindex - Trigger full, incremental, or file-specific reindexing
    4. configure_indexer - Runtime configuration updates
    5. clear_index - Reset and clear indexed data
    6. validate_config - Test connections and validate setup
  • Complete tool schemas with validation
  • Comprehensive error handling

  ✅ Phase 9: Testing & Documentation

  • Vitest testing infrastructure with 70% coverage thresholds
  • 58 passing tests across unit and integration suites
  • Comprehensive USAGE.md with examples and troubleshooting
  • Updated README with complete feature documentation
  • API documentation for all components

  📦 Phase 10: Release Preparation

  • Complete API.md with all component documentation
  • Changesets for version management
  • CHANGELOG.md initialization
  • npm publishing configuration
  • GitHub Actions workflows (CI, release, manual publish)
  • MIT license
  • Production-ready package.json

  🐛 Bug Fixes & Quality Improvements

  • Fixed all TypeScript compilation errors and warnings
  • Resolved 26 test failures - now 100% passing
  • Fixed ESLint configuration for proper linting
  • Addressed all 11 Copilot AI review comments:
    • Enhanced type safety (removed all any types from handlers)
    • Fixed baseUrl validation for local Ollama paths
    • Removed unused web-tree-sitter dependency
    • Added missing environment variable documentation
    • Improved error logging and signal handling
    • Added graceful shutdown with orchestrator cleanup

  📈 Current Status

  • ✅ 58 tests passing, 1 skipped
  • ✅ Zero compilation errors
  • ✅ Zero linting errors (35 non-blocking warnings)
  • ✅ Full TypeScript strict mode compliance
  • ✅ Comprehensive documentation
  • ✅ Ready for npm publishing

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog,
and this project adheres to Semantic Versioning.