gemini-file-search-cli

A CLI tool for managing Google Gemini File Search stores. Upload files, query with RAG (Retrieval Augmented Generation), and manage document metadata.

Features

• Store Management: Create, list, and delete File Search stores
• File Upload: Upload files and directories with automatic MIME type detection
• Mbox Support: Special handling for mbox email archives with metadata extraction
• Glob Patterns: Upload specific files from directories using glob patterns
• RAG Queries: Query your documents using natural language with Gemini models
• Metadata Filtering: Filter queries by custom metadata (year, author, etc.)

Installation

From npm

bun install -g gemini-file-search-cli
# or
npm install -g gemini-file-search-cli

From source

git clone https://github.com/ozanturksever/gemini-file-search-cli.git
cd gemini-file-search-cli
bun install
bun link

Requirements

• Bun runtime (v1.0.0+)
• Google Gemini API key

Setup

Set your Gemini API key as an environment variable:

export GEMINI_API_KEY=your_api_key_here

You can get an API key from Google AI Studio.

Usage

Store Management

# List all stores
gemini-file-search stores list

# Create a new store
gemini-file-search stores create my-docs

# Delete a store
gemini-file-search stores delete my-docs

Uploading Files

# Upload a single file
gemini-file-search upload ./document.pdf --store my-docs

# Upload a directory (recursive)
gemini-file-search upload ./docs --store my-docs

# Upload with glob pattern
gemini-file-search upload ./src --store my-code --glob "**/*.ts"

# Upload mbox email archive (extracts individual emails)
gemini-file-search upload ./archive.mbox --store my-emails

Managing Files

# List files in a store
gemini-file-search files list my-docs

# Delete files by metadata filter
gemini-file-search files delete my-docs --filter source_filename=old.pdf

Querying

# Basic query
gemini-file-search query my-docs "What are the main topics discussed?"

# Query with specific model
gemini-file-search query my-docs "Summarize the key points" --model gemini-2.5-pro

# Query with metadata filter
gemini-file-search query my-emails "emails about meetings" --filter year=2024

Supported File Types

The tool supports a wide range of file types including:

• Documents: PDF, Word, Excel, PowerPoint
• Code: TypeScript, JavaScript, Python, Java, and many more
• Text: Markdown, plain text, HTML, XML, JSON
• Email: mbox archives (with special processing)

See the Gemini File Search documentation for the full list.

Metadata

When uploading files, the tool automatically adds metadata:

• directory: The source directory path

For mbox files, additional metadata is extracted:

• from_email: Sender email address
• to_email: Recipient email addresses
• participants: All email participants (to, cc, bcc)
• year: Year the email was sent
• month: Month the email was sent
• source_filename: Original mbox filename

Models

Supported models for querying:

• gemini-3-pro-preview (default)
• gemini-2.5-flash
• gemini-2.5-flash-lite
• gemini-2.5-pro

License

MIT