Google has expanded the Gemini API File Search tool to support multimodal RAG, powered by Gemini Embedding 2. Images, charts, diagrams, PDFs, and code files can now be natively indexed in a unified semantic space alongside text. New features include page-level citations for verifiable answers, custom metadata filtering for targeted searches, and broad file format support (PDF, DOCX, Excel, CSV, JSON, Jupyter notebooks, PNG, JPEG). The infrastructure is fully managed — storage and query-time embeddings are free, with costs only for initial indexing and standard token usage. Developers can start building today using the Python SDK.
Table of contents
1. Multimodal RAG with Gemini Embedding 22. Enhanced Trust and Organization3. Managed Infrastructure and Pricing4. Broad File SupportSort: