Skip to main content

Integrations Overview

Vectorize integrates with external services through connectors to create powerful data pipelines for AI agents. These connectors work together to ingest diverse content sources, generate embeddings, and transform unstructured data into structured, metadata-rich context that agents can reason over.

How Connectors Enable Agentic AI​

Vectorize pipelines support retrieval and transformation workflows that enable agents to operate over structured context. Unlike traditional retrieval systems, agents built on Vectorize can:

  • Reason across multiple data sources to synthesize insights and make decisions
  • Use metadata for intelligent filtering - like prioritizing recent documents, specific authors, or document types
  • Perform multi-step analysis - such as comparing versions, detecting trends, or preparing comprehensive reports
  • Take action based on retrieved context - from generating summaries to triggering workflows

Example: Compliance Agent​

Instead of just retrieving policy documents, a compliance agent might:

  1. Analyze emails and documents for regulatory keywords
  2. Use metadata (sender, date, document type) to prioritize high-risk communications
  3. Cross-reference findings with policy documents
  4. Generate a compliance report with specific recommendations

Types of Connectors​

Source Connectors​

Source connectors allow you to ingest diverse content sources into structured pipelines, enabling agents to reason across emails, documents, databases, and more. They transform unstructured data from various sources into rich, queryable context.

Key Features:

  • Automatic metadata extraction (file type, creation date, author)
  • Support for various file formats with intelligent parsing
  • Configurable scheduling to keep agent knowledge current
  • OAuth integration for secure access to third-party services

Important: Once a source connector is associated with a pipeline, it cannot be removed or deleted.

AI Platform Connectors​

AI Platform connectors connect to AI services that generate embeddings and extract structured information from your content. They don't just create vector representations - they help transform raw text into structured data that agents can reason over.

Agent-Enabling Features:

  • Generate high-quality embeddings for semantic understanding
  • Extract structured metadata using schema-driven extraction
  • Support varies by model - some embedding providers support multiple languages or domains
  • Consistent model selection across ingestion and query time

Destination Connectors​

Destination connectors integrate with storage systems optimized for agent workloads. They don't just store vectors - they maintain structured, queryable indexes that agents can efficiently reason over.

Agent-Optimized Features:

  • Semantic search with metadata filtering for intelligent context selection
  • Namespace isolation for multi-tenant agent applications
  • Performance optimization for complex agent queries

Configuring Connectors for Agent Workflows​

You can configure connectors in two ways:

1. From the Integrations Section​

Navigate to the Vectorize dashboard and access each connector type from the left sidebar under "Integrations." This centralized approach allows you to:

  • Set up connectors that multiple agents can use
  • Configure metadata extraction rules
  • Test connector output before deploying to agents

2. While Creating a RAG Pipeline​

During pipeline creation, you can configure connectors in context:

  • See how source data flows through to agent-ready output
  • Configure metadata extraction specific to your use case
  • Preview how agents will interact with the processed data

Note: Connectors configured as part of a RAG Pipeline automatically appear in your organization's connector lists for reuse across other agent workflows.

Understanding Connector Reuse​

When you create a connector in Vectorize, you're creating a reusable integration with a service - for example, a Pinecone connector with your Pinecone API key, or a Fireflies connector with your Fireflies API key.

These connectors can be used across multiple pipelines. When you add a connector to a specific pipeline, you can configure additional parameters specific to that pipeline instance - like the Pinecone index name or a Fireflies meeting title filter.

This design lets you manage credentials securely while customizing connector behavior per pipeline.

Best Practices for Agentic AI​

  1. Design for reasoning, not just retrieval: Structure your data with rich metadata that agents can use for decision-making
  2. Think in workflows: Consider how agents will use multiple data sources together
  3. Optimize for quality: Better structured data leads to more capable agents
  4. Monitor and iterate: Track how agents use different data sources and refine accordingly
  5. Plan for scale: Choose connectors that can grow with your agent deployment

Getting Started​

Ready to build your first agent-powered application? Check out our guides:

For detailed configuration instructions, explore the individual connector pages in each category.

Was this page helpful?