Skip to content

v1.0.0 - Major Release: Functional Wrapper Generation

Choose a tag to compare

View all tags
@aberemia24 aberemia24 released this 22 Nov 17:16
· 16 commits to develop since this release
v1.0.0
07a0efd

v1.0.0 - Major Release: MCP Sampling & Refactor 🎉

This release marks a significant milestone with the stable introduction of LLM-in-the-loop (MCP) sampling capabilities, alongside a major refactor of the codebase and expanded multi-provider support.

🚀 New Features

MCP Sampling Implementation

  • ✨ Full LLM-in-the-loop sampling with hybrid MCP/API fallback architecture
  • 🔄 Multi-provider support: Anthropic, OpenAI, Gemini, Grok, Perplexity
  • 🐍 Python sampling interface with Pyodide
  • 📘 TypeScript sampling interface with SSE streaming
  • 📊 Audit logging, metadata, and Docker integration for sampling
  • 🤖 Multi-agent code review example (5 AI agents collaborating)

Configuration & Security

  • 📁 New config/ directory with sampling configuration schema
  • 🔒 Rate limiting implementation with AsyncLock protection
  • 🛡️ ContentFilter class for secret/PII detection and redaction
  • ✅ AJV deep schema validation for MCP tool calls

Developer Experience

  • 📂 New .agent/ folder with workflows and rules
  • 📚 Comprehensive sampling documentation (SAMPLING_SETUP.md, docs/sampling.md)
  • 💡 Working example: Multi-agent code review (examples/multi-agent-code-review.ts)

CLI Wizard Improvements

  • ✅ Wizard now fetches real tools from MCP servers (#71)
  • ✅ Functional wrappers with actual tool functions (not empty skeletons)

⚠️ Breaking Changes

Directory Restructuring

Core components reorganized for better maintainability:

  • config/ - Configuration and schemas
  • caching/ - Cache providers
  • core/ - Server, handlers, middleware
  • validation/ - Schema validation and security
  • executors/ - Sandbox executors
  • sampling/ - LLM sampling providers

Migration: Update import paths if you import from this package directly. Most users unaffected.

🐛 Bug Fixes

  • Fixed critical security and data integrity issues
  • Fixed hybrid sampling fallback (provider now created in MCP mode)
  • Fixed TypeScript errors and improved installer flow
  • Resolved test failures and deepened test coverage
  • Applied sampling environment variables in config discovery
  • Removed hardcoded Claude model, added multi-provider tests
  • Fixed CLI wizard empty wrapper generation (#71)

🔧 Configuration for Sampling

⚠️ IMPORTANT: Sampling currently works with API keys only (MCP SDK sampling coming soon).

Quick Setup:

  1. Create .env file:

    cp .env.example .env

  2. Add your API key:

    CODE_EXECUTOR_SAMPLING_ENABLED=true
CODE_EXECUTOR_AI_PROVIDER=gemini  # or anthropic, openai, grok, perplexity
GEMINI_API_KEY=your_key_here

  3. Use wrapper script (loads .env before starting):

    {
  "mcpServers": {
    "code-executor": {
      "command": "/path/to/start-with-env.sh"
    }
  }
}

See SAMPLING_SETUP.md for complete setup guide.

📦 Installation

npm

npm install -g code-executor-mcp@1.0.0

Docker

docker pull aberemia24/code-executor-mcp:1.0.0
# or
docker pull aberemia24/code-executor-mcp:latest

📊 Stats

  • +13,183 lines added
  • -420 lines removed
  • 133 files changed
  • Multi-provider sampling support
  • Enhanced security and validation

🔗 Links

This release provides a powerful and extensible platform for LLM-driven applications with autonomous multi-agent capabilities.

Assets 2
Loading