Gemini File Search Documents icon

Gemini File Search Documents

Manage and query documents in Gemini File Search stores

Actions7

Document → Query

AI-generated

Summary

Query documents in a Gemini File Search store using a specified model and prompt.

Inputs

  • model (required) — Gemini model to use (e.g., gemini-2.5-flash).
  • systemPrompt — System instructions that guide the model's behavior.
  • query (required) — Question or prompt to ask about the indexed documents.
  • storeName (required) — Identifier of the File Search store to query.
  • metadataFilter — AIP-160 filter expression to restrict documents.
  • includeSourceMetadata — When true, adds full document metadata to grounding chunks.

Output shape

An array of query result objects containing answer text and grounding information.

Each result includes $json.candidates[0].content.parts[0].text, grounding chunks, and optionally source metadata; errors are propagated unless continueOnFail is enabled.

Examples

Example 1: Answer a question about research findings.

model=gemini-2.5-flash, query='What are the key findings?', storeName=fileSearchStores/project_docs, includeSourceMetadata=true

Example 2: Retrieve metadata‑filtered results.

model=gemini-2.5-pro, query='List papers from 2023', metadataFilter='year=2023', storeName=fileSearchStores/papers, includeSourceMetadata=false

Example 3: Extract structured data using system prompt.

systemPrompt='Extract bullet points', query='Summarize the methodology', storeName=fileSearchStores/experiment_data, model=gemini-3-pro-preview, includeSourceMetadata=true

Discussion