Worked on the cloudwego/eino-examples repository to refactor the knowledge indexing orchestration data flow, focusing on improving reliability and scalability of the indexing pipeline. The approach involved restructuring the graph processing logic by introducing modular components such as FileLoader and MarkdownSplitter, ensuring that data flows sequentially through these stages before reaching the RedisIndexer. This change removed the previous direct edge from RedisIndexer to the end, clarifying component responsibilities and reducing system coupling. The work was implemented in Go, leveraging orchestration and graph processing skills to enhance maintainability and set the foundation for higher throughput in future indexing operations.