(Image credit: Google)
What you need to know
- Google is rolling out the File Search Tool, an easy-to-use RAG system that’s built into the Gemini API.
- The File Search Tool can leverage documents and databases to ground Gemini responses with factual context.
- Gemini API responses using the File Search Tool include citations that allow users to check the model’s work.
Google is adding a native retrieval-augmented generation (RAG) system to the Gemini API in an effort to make it easier for developers to back up their AI tools with hard data. It’s called the File Search Tool, and the company describes it as a “simple, integrated and scalable way to ground Gemini with…
(Image credit: Google)
What you need to know
- Google is rolling out the File Search Tool, an easy-to-use RAG system that’s built into the Gemini API.
- The File Search Tool can leverage documents and databases to ground Gemini responses with factual context.
- Gemini API responses using the File Search Tool include citations that allow users to check the model’s work.
Google is adding a native retrieval-augmented generation (RAG) system to the Gemini API in an effort to make it easier for developers to back up their AI tools with hard data. It’s called the File Search Tool, and the company describes it as a “simple, integrated and scalable way to ground Gemini with your data.” It results in outputs from the Gemini API that are “more accurate, relevant and verifiable,” per the announcement blog post.
The File Search Tool makes use of the Gemini Embedding model, but takes the heavy lifting required to develop a custom RAG system off the user. Instead, the tool handles file storage, chunking, embeddings, and dynamic context injection automatically. From there, the File Search Tool uses vector search to understand a given prompt and identify related information and data from provided documents.
Google says the File Search Tool can uncover relevant information from supplied documents even if the exact words or phrases aren’t used. Additionally, the model’s responses are easy to verify, as citations show exactly where the File Search Tool and Gemini API pulled an answer from. Users can then double check the model’s work to ensure accuracy.
The File Search Tool supports a wide variety of application and text-based file formats, listed here. It includes most major file formats and programming types, like PDF, DOCX, TXT, and JSON. This makes it easy to take your existing database and hand it to the Gemini API via the File Search Tool.
What you need to try the File Search Tool
There’s a demo app in the Google AI Studio that gives developers an idea of how the File Search Tool can be used in their business applications. The tool is paid, but developers only need to pay a fixed rate of $0.15 per 1 million tokens when initially embedding and indexing their files. After that, storage and embedding generation is free for each individual query.
Google says this payment model “makes the File Search Tool both significantly easier and very cost-effective to build and scale with.” It’s available now in the Gemini API for those that want to try it out.
(Image credit: Google)
This is the second Gemini API announcement of the week, as Google previously unveiled support for JSON Schema, which makes it easier to use the API in multi-agent workflows.
Get the latest news from Android Central, your trusted companion in the world of Android
Brady is a tech journalist for Android Central, with a focus on news, phones, tablets, audio, wearables, and software. He has spent the last three years reporting and commenting on all things related to consumer technology for various publications. Brady graduated from St. John’s University with a bachelor’s degree in journalism. His work has been published in XDA, Android Police, Tech Advisor, iMore, Screen Rant, and Android Headlines. When he isn’t experimenting with the latest tech, you can find Brady running or watching Big East basketball.