Developed a document ingestion enhancement for the NousResearch/hermes-agent repository, focusing on improving the reliability of text-based document uploads. Addressed the challenge of unreliable or missing MIME types by implementing a fallback mechanism that leverages file extensions for accurate classification. This approach ensures that various text formats are correctly processed and injected into the agent’s context, reducing ingestion errors and strengthening downstream data processing. The work centered on backend development and API integration, utilizing Python to refine document handling for Matrix uploads. The solution increased data fidelity and resilience, supporting more robust search and processing capabilities within the Hermes agent system.