Source: File Search (Gemini API docs)
Google's hosted RAG tool for Gemini: persistent file search stores, chunking config, metadata filters, file_citation annotations, limits and pricing.
Contents
- What: official API documentation page
- Publisher: Google (Gemini API / AI Studio)
- Date: living document; read 2026-11-14
- Original: File Search
Summary
File Search imports, chunks, embeds and indexes documents in a persistent store that the file_search tool queries during an interaction.
Key claims
- Stores persist until deleted (unlike raw Files, which expire after 48 hours); embeddings have no TTL.
- Ingest by
uploadToFileSearchStoreor by Files API upload plusimportFile; both are long-running operations. - Optional
chunking_config.white_space_config(max_tokens_per_chunk,max_overlap_tokens). - Custom metadata plus AIP-160
metadata_filter; citations come back asfile_citationannotations with file name, page number and metadata. - Image indexing with
gemini-embedding-2. - Limits: 100 MB per document; store size 1 GB (Free) to 1 TB (Tier 3); keep stores under 20 GB for latency; backend size is about 3x input.
- Pricing: pay embedding rates at indexing; storage and query embeddings are free; retrieved chunks are billed as context tokens.
- Cannot currently be combined with Google Search grounding or URL context; can be combined with custom functions on Gemini 3/4.