Multi-Source Documentation MCP Server

""" Multi-Source Documentation MCP Server Provides 6 specialized tools for searching Voiceflow and Claude Code documentation. """ from fastmcp import FastMCP from pathlib import Path import sys from .parser import VoiceflowDocsParser from .db_manager import VoiceflowDocsDB from .config import Config # Initialize MCP server mcp = FastMCP("Multi-Source Documentation") # Global instances (initialized on startup) config = None db = None parser = None def initialize_server(): """Initialize database with documentation from all sources""" global config, db, parser print("🚀 Initializing Multi-Source Documentation MCP Server...") # Load configuration try: config = Config.from_env() config.validate() print(f" Documentation sources: {', '.join(config.docs_paths.keys())}") for source, path in config.docs_paths.items(): print(f" - {source}: {path}") print(f" Database path: {config.db_path}") except Exception as e: print(f"❌ Configuration error: {e}") sys.exit(1) # Initialize database db = VoiceflowDocsDB(config.db_path) parser = VoiceflowDocsParser() # Check if database needs indexing stats = db.get_stats() if stats['total_documents'] == 0: print(" Database is empty, indexing documentation...") all_documents = [] # Parse all documentation sources for source, docs_path in config.docs_paths.items(): print(f" Parsing {source} documentation from {docs_path}...") documents = parser.parse_directory(docs_path) # Ensure source is set for doc in documents: doc['source'] = source print(f" Found {len(documents)} {source} documents") all_documents.extend(documents) # Index in database print(f" Indexing {len(all_documents)} total documents in database...") indexed = db.index_documents(all_documents) print(f" Indexed {indexed} documents") # Show stats by source stats = db.get_stats() if 'by_source' in stats: print(f" Documents by source:") for source, count in stats['by_source'].items(): print(f" - {source}: {count} docs") else: print(f" Database already indexed: {stats['total_documents']} documents, {stats['total_code_examples']} code examples") if 'by_source' in stats: print(f" Documents by source:") for source, count in stats['by_source'].items(): print(f" - {source}: {count} docs") print("✅ Server ready!") @mcp.tool() def search_documentation(query: str, source: str = "all", category: str = None, limit: int = 5) -> dict: """ Search Voiceflow and Claude Code documentation with full-text search. Use this tool to find documentation pages about features, integrations, or concepts. Args: query: Search query (e.g., "MCP integration", "KB search step", "custom actions") source: Documentation source filter - "voiceflow", "claude-code", or "all" (default: "all") category: Optional category filter limit: Maximum number of results to return (default: 5) Returns: Dictionary with search results including title, URL, snippet, source, and document ID Example: search_documentation("how to use MCP tool") search_documentation("API step configuration", source="voiceflow", limit=3) search_documentation("Claude Code setup", source="claude-code") """ if not db: return {"error": "Database not initialized"} try: results = db.search(query, source=source, category=category, limit=limit) return { "query": query, "source_filter": source, "count": len(results), "results": results } except Exception as e: return {"error": str(e)} @mcp.tool() def get_document(doc_id: int) -> dict: """ Retrieve complete document content by ID. Use this tool to get the full documentation content after finding a document with search_documentation. Args: doc_id: Document ID from search results Returns: Full document with title, content, URL, code examples, and metadata Example: get_document(42) """ if not db: return {"error": "Database not initialized"} try: doc = db.get_document(doc_id) if not doc: return {"error": f"Document ID {doc_id} not found"} return doc except Exception as e: return {"error": str(e)} @mcp.tool() def search_code_examples(query: str, language: str = None, limit: int = 5) -> dict: """ Search for code examples in Voiceflow documentation. Use this tool to find code snippets and implementation examples. Args: query: What kind of code to find (e.g., "MCP tool definition", "API request", "webhook") language: Optional language filter (e.g., "javascript", "python", "json", "typescript") limit: Maximum number of code examples to return (default: 5) Returns: Dictionary with code examples, their language, source document, and URL Example: search_code_examples("MCP tool configuration") search_code_examples("API request", language="json") """ if not db: return {"error": "Database not initialized"} try: examples = db.search_code_examples(query, language=language, limit=limit) return { "query": query, "language": language, "count": len(examples), "examples": examples } except Exception as e: return {"error": str(e)} @mcp.tool() def list_categories(source: str = "all") -> dict: """ List all documentation categories with document counts by source. Use this tool to see what documentation categories are available. Args: source: Documentation source filter - "voiceflow", "claude-code", or "all" (default: "all") Returns: Dictionary with category list organized by source and total document count Example: list_categories() list_categories(source="voiceflow") list_categories(source="claude-code") """ if not db: return {"error": "Database not initialized"} try: categories = db.list_categories(source=source) stats = db.get_stats() return { "source_filter": source, "categories": categories, "total_docs": stats['total_documents'], "total_code_examples": stats['total_code_examples'], "by_source": stats.get('by_source', {}) } except Exception as e: return {"error": str(e)} @mcp.tool() def get_integration_docs(integration: str) -> dict: """ Get documentation for a specific Voiceflow integration. Use this tool to find docs about integrations like MCP, Zapier, Make, Airtable, etc. Args: integration: Integration name (e.g., "mcp", "zapier", "make", "hubspot", "salesforce") Returns: Main documentation and related docs for the integration Example: get_integration_docs("mcp") get_integration_docs("zapier") """ if not db: return {"error": "Database not initialized"} try: # Search for integration results = db.search(f"{integration} integration tool", limit=3) if not results: return {"error": f"No documentation found for '{integration}' integration"} # Get full content for first result main_doc = db.get_document(results[0]['id']) return { "integration": integration, "main_doc": main_doc, "related_docs": results[1:] if len(results) > 1 else [] } except Exception as e: return {"error": str(e)} @mcp.tool() def find_step_documentation(step_name: str) -> dict: """ Find documentation for a specific Voiceflow step. Use this tool to get docs about Voiceflow steps like API, Prompt, KB Search, Function, etc. Args: step_name: Step name (e.g., "API", "Prompt", "KB Search", "Function", "Custom Action", "Agent") Returns: Main documentation for the step and related pages Example: find_step_documentation("API") find_step_documentation("KB Search") find_step_documentation("Prompt") """ if not db: return {"error": "Database not initialized"} try: # Search for step documentation results = db.search(f"{step_name} step", limit=3) if not results: return {"error": f"No documentation found for '{step_name}' step"} # Get full content for first result main_doc = db.get_document(results[0]['id']) return { "step_name": step_name, "documentation": main_doc, "related": results[1:] if len(results) > 1 else [] } except Exception as e: return {"error": str(e)} def main(): """Main entry point for MCP server""" # Initialize on startup initialize_server() # Run MCP server mcp.run() if __name__ == "__main__": main()