WeWork MCP Server

Instructions

Implementation Reference

• wework_mcp_server.py:204-254 (handler)

  The main handler function for the 'find_project_by_name' tool, decorated with @mcp.tool() for automatic registration and schema inference. It fetches all projects and uses similarity matching to find the best match.
  @mcp.tool() def find_project_by_name(project_name: str, threshold: float = 0.3) -> Dict[str, Any]: """ Tìm dự án theo tên với độ tương đồng Args: project_name: Tên dự án cần tìm threshold: Ngưỡng tương đồng tối thiểu (default: 0.3) Returns: Thông tin dự án phù hợp nhất """ try: if not wework_client: return {'error': 'WeWork client not initialized'} logger.info(f"Finding project by name: {project_name}") projects = wework_client.fetch_projects() if not projects: return { 'error': 'No projects available or failed to fetch projects', 'success': False } best_project, similarity_score = wework_client.find_best_project_match( project_name, projects, threshold ) if best_project: return { 'success': True, 'found': True, 'project': best_project, 'similarity_score': similarity_score, 'search_term': project_name } else: return { 'success': True, 'found': False, 'similarity_score': similarity_score, 'search_term': project_name, 'message': f'Không tìm thấy dự án phù hợp với "{project_name}" (ngưỡng: {threshold})', 'available_projects': [p.get('name', 'Unknown') for p in projects[:5]] # Show first 5 for reference } except Exception as e: logger.error(f"Error in find_project_by_name: {e}") return {'error': str(e), 'success': False}
• wework_mcp_server.py:205-215 (schema)

  Input/output schema defined by type hints (project_name: str, threshold: float=0.3 -> Dict[str, Any]) and docstring describing parameters and return value.
  def find_project_by_name(project_name: str, threshold: float = 0.3) -> Dict[str, Any]: """ Tìm dự án theo tên với độ tương đồng Args: project_name: Tên dự án cần tìm threshold: Ngưỡng tương đồng tối thiểu (default: 0.3) Returns: Thông tin dự án phù hợp nhất """
• data/wework_client.py:212-269 (helper)

  Supporting utility method in WeWorkClient that implements the fuzzy matching logic using TF-IDF cosine similarity (sklearn) or fallback set-based string similarity.
  def find_best_project_match(self, target_name: str, projects: List[Dict], threshold: float = 0.3) -> Tuple[Optional[Dict], float]: """ Tìm dự án phù hợp nhất bằng cosine similarity (nếu sklearn có sẵn) hoặc simple string matching """ if not projects: return None, 0 # Tạo danh sách tên dự án project_names = [project['name'].lower() for project in projects] target_name_lower = target_name.lower() # Kiểm tra khớp chính xác trước for i, name in enumerate(project_names): if target_name_lower in name or name in target_name_lower: return projects[i], 1.0 # Sử dụng TF-IDF và cosine similarity nếu sklearn có sẵn if SKLEARN_AVAILABLE: try: vectorizer = TfidfVectorizer(analyzer='char', ngram_range=(2, 3)) all_names = project_names + [target_name_lower] tfidf_matrix = vectorizer.fit_transform(all_names) # Tính cosine similarity giữa target và tất cả project names target_vector = tfidf_matrix[-1] project_vectors = tfidf_matrix[:-1] similarities = cosine_similarity(target_vector, project_vectors).flatten() # Tìm similarity cao nhất best_idx = np.argmax(similarities) best_similarity = similarities[best_idx] if best_similarity >= threshold: return projects[best_idx], best_similarity else: return None, best_similarity except Exception as e: print(f"Lỗi khi tính cosine similarity: {e}") # Fallback: Simple string similarity best_match = None best_score = 0 for project in projects: name_lower = project['name'].lower() # Simple similarity based on common characters common_chars = len(set(target_name_lower) & set(name_lower)) total_chars = len(set(target_name_lower) | set(name_lower)) similarity = common_chars / total_chars if total_chars > 0 else 0 if similarity > best_score and similarity >= threshold: best_match = project best_score = similarity return best_match, best_score