Worked on the 0xPlaygrounds/rig repository to deliver support for Retrieval-Augmented Generation (RAG) documents within the Gemini API, enabling agents to dynamically utilize context from vector stores. The implementation involved modifying the request body to inject documents as the initial user messages and converting TXT files to plain text parts while preserving other document types. Comprehensive tests were developed to validate document handling and ensure backward compatibility with existing workflows. This work was accomplished using Rust, with a focus on API development and robust testing practices, resulting in enhanced agent capabilities for dynamic context retrieval without disrupting established document processing.