FoulSpark/Advanced-MCP-Server-for-Intelligent-Code-Completion

Advanced MCP Server for Intelligent Code Completion

A local Flask-based Model Context Protocol (MCP) server that powers intelligent, context-aware code completion and automated fixing workflows for your editor (with a bundled VS Code extension). It indexes your project, builds a symbol/dependency graph, performs semantic search, integrates with Git and GitHub/GitLab via MCP or direct APIs, and can run a self‑healing loop to plan and apply fixes using Gemini.

🚀 Key Features

• Project understanding
  • Project-wide indexing (functions/classes/imports/variables) via AST and patterns
  • Symbol/dependency graph across files
  • Lightweight semantic search for relevant symbols and files
• Intelligent assistance
  • Code analysis with error detection (AST + Pyflakes)
  • Pattern detection for common syntax issues (e.g., missing colons) and undefined variables
  • Optional formatters: autopep8, black, isort
• MCP + Git integration
  • Enhanced MCP manager with fallbacks to direct REST where needed
  • GitHub/GitLab MCP servers (via npx @modelcontextprotocol/server-*) and filesystem MCP
  • Local git history search for example-driven context
• Self-healing fixer (optional)
  • Multi-iteration loop that analyzes issues, plans fixes with Gemini, and applies changes
  • Keeps a simple learning DB of successful fixes and trends
• VS Code extension (SuperAI)
  • Four-stage Gemini workflow (understand → plan → implement → validate/replace)
  • Commands for diagnostics, completions, and enhanced syntax fixes

🗂️ Repository Structure

• advanced_mcp_server.py — Flask app, API key rotation, code analysis, optional watchers
• enhanced_mcp_manager.py — MCP client wrapper with GitHub/GitLab/filesystem servers and fallbacks
• mcp_git_integration.py — Legacy/alternate MCP Git integration
• project_indexer.py — Project scanner and indexer
• semantic_search.py — Simple embedding-based search engine
• symbol_graph.py — Dependency graph between files/symbols (NetworkX)
• multi_file_manager.py — Orchestrates indexing, semantic search, and refactoring helpers
• multi_file_editor.py — Generates/apply multi-file patches with diffs and previews
• self_healing_code_fixer.py — Optional Gemini-powered self-healing loop
• check_gemini_models.py — Utility to list/validate available Gemini models for a key
• superai-extension/ — VS Code extension (README, extension.js, etc.)
• start_enhanced_mcp.bat / start_enhanced_mcp.sh — Helper scripts
• .env.template — Example environment configuration
• requirements.txt / requirements_self_healing.txt — Python dependencies

⚙️ Requirements

• Python 3.10+
• Node.js (for MCP servers via npx if you use GitHub/GitLab/filesystem MCP)
• Git (for local history and repo operations)

Python dependencies are listed in:

• requirements.txt (core server)
• requirements_self_healing.txt (optional self-healing loop)

Install with:

pip install -r requirements.txt
# Optional for self-healing
pip install -r requirements_self_healing.txt

🔐 Configuration

Copy .env.template to .env and fill values:

• GEMINI_API_KEY (required)
• GEMINI_API_KEY_1..9, GEMINI_API_KEY_STAGE0/1/2 (optional rotation pool)
• GITHUB_TOKEN, GITLAB_TOKEN (optional, for MCP servers or API fallbacks)
• MCP_SERVER_URL, WORKSPACE_ROOT

The server includes a smart API key rotation that automatically switches keys after consecutive failures.

Note: Avoid hardcoding API keys in code. Prefer environment variables.

▶️ Run the Server

Windows:

start_enhanced_mcp.bat

macOS/Linux:

./start_enhanced_mcp.sh

Or directly:

python advanced_mcp_server.py

By default the server listens on http://127.0.0.1:5000 (as used by the VS Code extension).

🧩 VS Code Extension (SuperAI)

The superai-extension/ folder contains a VS Code extension that connects to this server.

Highlights:

• Four-stage Gemini pipeline: Understand → Plan → Implement → Validate/Replace
• Enhanced syntax fix (missing colons, undefined vars, auto-create functions)
• API key monitoring, diagnostics, completions, and global fixes

See superai-extension/README.md for detailed usage, commands, and troubleshooting.

🤖 Self-Healing Loop (Optional)

self_healing_code_fixer.py can run a loop that:

• Analyzes a file via the MCP server
• Plans enhanced fixes with Gemini
• Applies patches and records learning signals

Environment:

• Requires google-generativeai and a valid GEMINI_API_KEY

Example invocation (see file for demo):

python self_healing_code_fixer.py

🔍 How It Works (High Level)

• Indexing and Search

  • ProjectIndexer parses Python via AST and uses patterns for JS/TS and other languages
  • SemanticSearchEngine builds simple embeddings and supports concept search
  • SymbolGraph builds relationships (imports/contains/calls/uses) for impact analysis

• MCP/Git Context

  • EnhancedMCPManager tries MCP servers first (GitHub/GitLab/filesystem), then falls back to direct APIs
  • Local git history excerpts can be surfaced to guide completions/refactors

• Analysis and Fixing

  • advanced_mcp_server.py performs file analysis, optional linting (pyflakes), and formatting (autopep8/black/isort)
  • Pattern checks catch common syntax/variable issues before AST parse

🧪 Utilities

• check_gemini_models.py lists available Gemini models for a key and can perform tiny validation calls
  • Prefer environment variables; remove any hardcoded keys before committing

❗ Troubleshooting

• Ensure .env has a valid GEMINI_API_KEY
• If MCP GitHub/GitLab servers fail, fallbacks to direct APIs may be used (requires tokens)
• If embeddings or graph features are slow on very large repos, start with a smaller subset
• On Windows, use the .bat launcher; on Unix, the .sh script

🙌 Acknowledgements

• Model Context Protocol (MCP)
• Gemini API
• NetworkX

Generated README based on the current repository content to help you publish on GitHub quickly.