Google Upgrades Gemini API File Search
Google has introduced three major upgrades to the File Search tool within its Gemini API. The updates add multimodal retrieval, metadata filtering, and page-level citations to the platform’s retrieval-augmented generation system.
The biggest enhancement enables developers to search images and text documents within the same File Search environment. Powered by the Gemini Embedding 2 model, the system can index charts, diagrams, product photos, and other visual content without OCR processing.
As a result, developers can build applications that support more advanced visual retrieval workflows. Meanwhile, teams focused on text-only operations can continue using the gemini-embedding-001 model for lower-cost processing.
Metadata Filtering Improves Retrieval Accuracy
Google has also introduced custom metadata filtering for document organization and targeted search. Developers can attach labels such as category, season, or pricing tier during uploads. They can then apply filter expressions to narrow searches within specific document groups.
Consequently, the update addresses a major challenge in enterprise RAG systems where different document types often share the same storage environment. The filtering system, therefore, improves contextual relevance and retrieval precision.
In addition, File Search responses now include page-level citations. The platform identifies exact document pages linked to retrieved information. Furthermore, multimodal stores can generate downloadable image references through dedicated media identifiers.
Enterprise RAG Features Expand
Google first launched File Search in public preview during November 2025 as a managed RAG solution integrated into the generateContent API. The platform handles document chunking, indexing, embedding, and retrieval through a unified workflow.
The latest updates also support structured output with Gemini 3 models and configurable chunking strategies. Therefore, developers gain greater control over how systems split and retrieve documents.
Google highlighted several enterprise use cases, including insurance claims processing, visual product search, research documentation, and design system libraries. As enterprise AI adoption accelerates, the upgraded File Search platform aims to support more scalable and production-ready retrieval workflows.
