Skip to content

Latest commit

 

History

History
164 lines (125 loc) · 5.14 KB

File metadata and controls

164 lines (125 loc) · 5.14 KB

Changelog

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.

Added

  • Production hardening improvements (memory management, rate limiting, health checks)

0.1.0 - 2025-11-26

Added

Core Features

  • MCP Server with 5 specialized adapters for AI tool integration
    • dev_search - Semantic code search with type-aware understanding
    • dev_status - Repository health and statistics
    • dev_plan - Implementation planning from GitHub issues
    • dev_inspect - File analysis (similarity + pattern checking)
    • dev_gh - GitHub issue/PR search with offline caching
  • Multi-language Support - TypeScript, JavaScript, Go, Python, Rust, Markdown
  • Local-first Architecture - All embeddings and indexing run locally
  • Subagent System - Coordinator with Explorer, Planner, and PR agents
  • Event Bus - Async pub/sub communication between components
  • Observability - Request tracking, structured logging, p50/p95/p99 metrics

Developer Experience

  • CLI - Complete command-line interface with Commander.js
  • Cursor Integration - One-command setup (dev mcp install --cursor)
  • Claude Code Integration - MCP server configuration for Claude Desktop
  • Centralized Storage - Indexes stored globally in ~/.dev-agent/indexes/
  • Auto-reload - GitHub index automatically reloads on changes
  • Token Estimation - Real-time cost tracking (<1% error rate)

Infrastructure

  • Circular Buffers - Prevent memory leaks in long-running processes
  • Type Safety - TypeScript strict mode enabled throughout
  • Testing - 1012 tests passing with comprehensive coverage
  • CI/CD - GitHub Actions for linting, testing, and releases
  • Monorepo - Turborepo + pnpm workspaces for efficient builds

Changed

  • Message history uses circular buffer (max 1000 messages)
  • Response times use circular buffer (max 1000 entries)
  • Improved error messages with actionable suggestions

Fixed

  • Memory leaks from unbounded array growth
  • Zombie MCP server processes when Cursor closes
  • GitHub index not reloading after dev github index
  • STDIO transport not handling stdin closure properly

Security

  • Input validation on all MCP tool adapters
  • Rate limiting per tool (100 requests/minute)
  • Graceful error handling without exposing internals
  • Memory bounds to prevent resource exhaustion
  • No telemetry or cloud dependencies - fully local

Performance

  • Response times < 2s for most operations
  • Memory-bounded data structures prevent leaks
  • Lazy loading for GitHub indexer
  • Efficient vector search with LanceDB

Documentation

  • Comprehensive README with examples
  • Architecture documentation (ARCHITECTURE.md)
  • Contribution guidelines (CONTRIBUTING.md)
  • Agent and workflow documentation (AGENTS.md, CLAUDE.md)
  • MCP setup guides for Cursor and Claude Code

Known Limitations

  • STDIO transport only (HTTP planned for v0.2.0)
  • Sequential tool execution (parallel planned for v0.2.0)
  • No built-in caching layer (planned for v0.2.0)
  • English language content optimized (multilingual improvements planned)

0.0.1 - 2025-11-20

Added

  • Initial development setup
  • Core repository scanner with ts-morph
  • Vector storage with LanceDB
  • Basic CLI commands

Upgrade Guides

Upgrading to 0.1.0

From Development Versions:

  1. Install globally:

    npm install -g dev-agent
  2. Re-index your repositories:

    cd /path/to/repo
    dev index .
    dev github index  # If using GitHub integration
  3. Reinstall MCP integration:

    dev mcp install --cursor
    # or
    dev mcp install  # for Claude Code
  4. Restart your IDE (Cursor/Claude Desktop)

Breaking Changes:

  • None (first public release)

Deprecations:

  • None

Release Notes

v0.1.0 Release Highlights

This is the first public release of dev-agent! 🎉

What makes dev-agent special:

  • 100% Local - No API keys, no cloud dependencies
  • Deep Understanding - AST-based analysis + semantic search
  • Action-Capable - Not just search, but planning and automation
  • MCP-Native - First-class integration with Claude and Cursor
  • Production-Grade - Memory-bounded, rate-limited, well-tested

Use Cases:

  • Semantic code search across large codebases
  • Implementation planning from GitHub issues
  • Code pattern discovery and analysis
  • Offline GitHub issue/PR search
  • Repository health monitoring

Next Steps:

  • Try the examples in examples/
  • Read the setup guides for Cursor or Claude Code
  • Join discussions and provide feedback
  • Contribute! See CONTRIBUTING.md

Support