Worked on the deepset-ai/haystack-core-integrations repository to enhance the MCP Tool by enabling reliable extraction of the first text element from mixed content types. This involved a focused refactor in Python to reduce code duplication and simplify ongoing maintenance, while preserving existing behavior. Expanded automated test coverage to validate extraction logic across various content types and improved CI reliability by conditionally skipping tests when the OPENAI_API_KEY is absent. Emphasized backend development and API integration, with additional improvements to documentation, type hints, and fallback handling. The result was a more robust content ingestion pipeline and clearer ownership of extraction logic.