Vector Stores Quickstart

Transform your documents into intelligent search in under 5 minutes.

This quickstart covers both the basics and key advanced features: creating Vector Stores, uploading files with metadata, performing semantic search, and using filters.

Prerequisites

Before you begin, make sure you have:

API Key: Get your API key from the API Keys page
SDK Installed: Install the Mixedbread SDK for your preferred language

pip install mixedbread

Step 1: Create Your First Vector Store

A Vector Store is a container that organizes your files for semantic search. Think of it as a smart folder that understands the content and meaning of your documents.

The response contains a complete Vector Store object with a unique ID. You can use either the ID or name as identifiers in all operations.

For complete details on Vector Store configuration including expiration policies and public access, see Creating Vector Stores.

Step 2: Upload and Process Files

Upload files to your Vector Store. The system automatically extracts content, creates semantic chunks, generates embeddings, and indexes everything for fast retrieval.

Supported File Types: We support documents, presentations, code, images, and specialized formats. See Supported File Types for the complete list.

File Processing: Files progress through processing stages from upload to being searchable. For details on status tracking, see File Ingestion.

# Upload with metadata
file_response = mxbai.vector_stores.files.upload(
    vector_store_id="my-knowledge-base",
    file=open("product-guide.pdf", "rb"),
    metadata={
        "category": "documentation",
        "department": "product", 
        "version": "2.1",
        "last_updated": "2024-01-15"
    }
)

Search with Metadata Filtering

Combine semantic search with precise filtering:

# Search with metadata filters
results = mxbai.vector_stores.search(
    query="How does authentication work?",
    vector_store_identifiers=["my-knowledge-base"],
    metadata_filter={
        "category": {"$eq": "documentation"},
        "version": {"$gte": "2.0"}
    }
)

Vector Store Configuration

Create Vector Stores with custom settings:

# Vector Store with expiration and public access
vector_store = mxbai.vector_stores.create(
    name="public-docs",
    is_public=True,  # Allow other users to search
    expires_after={
        "anchor": "last_active_at",
        "days": 30  # Auto-delete after 30 days of inactivity
    }
)