Worked on enhancing the embedding capabilities and robustness of the text processing pipeline in the GraphScope/portal repository. Focused on backend development using Python, the work involved integrating an embedding model and refactoring the DefaultEmbedding component to inherit from EmbeddingFunction. This unified embedding usage across portal components and aligned the workflow with ChromaDB integration, improving retrieval accuracy and consistency. The refactor streamlined the embedding process, reduced configuration drift, and simplified future maintenance. By increasing the robustness of the pipeline and reducing edge-case failures, the changes laid a foundation for easier model swaps and more reliable natural language processing workflows.