Google drops managed RAG tool inside Gemini API
Google just launched File Search, a fully managed RAG pipeline integrated directly into the Gemini API. The tool automates file storage, chunking, vector retrieval, and automated source citations. Google is also waiving storage and query-time embedding fees, charging only a one-time indexing fee of $0.15 per million tokens. Developers can now ground Gemini models in custom documents without wrestling with custom vector databases.
Building a custom knowledge base for your AI apps usually means dealing with an endless headache of document chunking, vector databases, and messy retrieval pipelines.
Google wants to wipe all of that off your plate.
Google just introduced the File Search Tool in the Gemini API. It is a fully managed Retrieval-Augmented Generation (RAG) system baked right into the platform, handling the heavy lifting behind the scenes so developers can focus on actually building.
Free storage and zero query embedding fees
The most aggressive part of this launch is definitely the pricing model.
Google is making both document storage and embedding generation at query time completely free of charge. You only pay when you first index your documents, charged at $0.15 per 1 million tokens using the gemini-embedding-001 model.
Under the hood, the tool takes care of optimal chunking strategies, vector embeddings, and dynamically injecting retrieved context directly into prompts via the standard `generateContent` API.
It handles a wide range of formats out of the box, including PDF, DOCX, TXT, JSON, and common programming language files. Best of all, responses automatically include verifiable citations pointing directly to the source documents.
Early testers are already putting it through its paces. Phaser Studio plugged File Search into Beam, its AI-powered game generation tool. The team runs thousands of daily queries across a massive template library, pulling parallel cross-corpus results in under two seconds instead of spending hours manually cross-referencing.
Developers can test this immediately via the Ask the Manual demo in Google AI Studio with a paid API key.
Why it matters
We think Google is taking direct aim at OpenAI Assistants API, and it is using aggressive pricing as its main weapon.
For solo builders and startups, maintaining dedicated vector databases like Pinecone can get expensive fast. By offering free file storage and waiving search-time embedding fees, Google is removing the operational friction of building AI support bots and internal search engines.
It makes standing up a functional RAG pipeline practically free to experiment with. If your document sets stay relatively static, ongoing operational costs basically drop to just the Gemini model generation tokens.
That said, full abstraction is always a double-edged sword. You lose fine-grained control over chunking strategies, hybrid search tuning, and reranking parameters. For messy or highly technical documents, that lack of customization might hurt retrieval accuracy. But for standard business docs, this feels like an absolute no-brainer.
Sources (1)
- blog.google(opens in a new tab)Originalblog.google