Structures Explained

Structures in Gen AI Builder are Griptape-powered agents, pipelines, and workflows that enable advanced AI logic, data processing, and system integration.

Structures are foundational to building production-ready, intelligent applications with EDB Postgres® AI (EDB PG AI):

• Execute complex AI workflows
• Transform and enrich data
• Implement custom Tools and Retrievers
• Integrate with external APIs and business systems

Without Structures, Assistants and RAG pipelines are limited to static retrieval and generation — Structures provide the custom logic that brings intelligence and business context to your AI Factory applications.

Before you start

Prerequisites for understanding Structures:

• Familiarity with Assistants and Knowledge Bases
• Awareness of Retrievers, Rulesets, and Tools
• Basic understanding of Griptape Structures and RAG pipelines

Suggested reading:

• Assistants Explained
• Rulesets Explained
• Retrievers Explained
• Knowledge Bases Explained
• AI Factory Concepts

What is a Structure?

A Structure is a reusable component that encapsulates advanced AI behavior:

• AI agents and pipelines
• Data transformations
• API integrations
• Workflow orchestration
• Custom retrieval logic

Structures can be:

• Run on demand from the AI Factory Web Console
• Invoked programmatically via API
• Published as Tools for Assistants
• Integrated into Data Source pipelines and Knowledge Base pipelines

In short: Structures bring custom intelligence and business process integration to your AI Factory applications.

Why use Structures?

• Define organization-specific business logic
• Support multi-step reasoning and workflow orchestration
• Extend Assistants with custom Tools
• Enrich and transform incoming data
• Implement custom retrieval pipelines
• Integrate AI with external systems (APIs, databases, SaaS platforms)

Structures make your AI applications extensible, compliant, and deeply integrated with enterprise systems.

How Structures work

Structures are implemented as:

• Griptape Agents → conversational AI logic
• Griptape Pipelines → multi-step data transformation
• Griptape Workflows → complex orchestration across agents and tools

A Structure includes:

• Structure config → YAML or Python definition
• Structure code → Python, packaged as Zip or GitHub repo
• Environment variables → for API keys, secrets, and parameters

Execution paths:

API Call → Structure Execution
Assistant → Tool → Structure Execution
Data Source Pipeline → Structure Execution

Where Structures fit in AI Factory pipelines

Structures can:

• Implement custom Data Sources
• Perform Data Source transformations
• Provide Tools for Assistants
• Enrich RAG pipelines with custom retrieval logic

When to use Structures

Use Structures when you need:

• Advanced AI capabilities beyond simple Q&A
• Complex data transformations
• Business workflow orchestration
• API integrations and external lookups
• Dynamic Tools for Assistants
• Custom search with advanced ranking or filtering

Typical use cases:

• Transaction categorization and tagging
• Data anonymization for privacy compliance
• Product catalog enrichment
• API-driven Tools for Assistants (inventory, pricing, logistics)
• Complex summarization agents
• Custom hybrid retrieval logic for RAG pipelines

Patterns of use

Standalone Structure

• Executed on demand via UI or API
• Example: Data Anonymizer run manually or on schedule

Structure as a Tool

• Published as a Tool in AI Factory
• Invoked by Assistants at runtime
• Example: Product Inventory Lookup Tool used by Support Assistant

Structure as Data Transformer

• Integrated into Data Source pipeline
• Transforms raw data before indexing
• Example: PDF Parser + Metadata Enricher pipeline

Structure as Custom Retriever / Data Source

• Implements a custom Retriever or custom Data Source
• Example: Real-Time Pricing Retriever that queries external pricing APIs and merges results with Knowledge Base content

Structure as Business Workflow Orchestrator

• Executes multi-step business logic
• Example: Order Processing Workflow → fetch inventory, calculate shipping, update CRM, notify customer

Best practices

• Package Structures cleanly:
• Clear Structure Config
• Minimal, isolated external dependencies
• Externalize secrets via environment variables
• Implement robust error handling and retries
• Validate performance for production use cases
• Use versioned Data Lake paths where applicable
• Monitor execution performance and success rates
• Follow Griptape best practices:
• Modular pipelines and workflows
• Stateless agents where possible
• Clear separation of transformation and orchestration logic
• Use audit trails for critical business workflows
• Maintain governance for any externally integrated APIs or data

Sovereign AI alignment

Structures are a key enabler of Sovereign AI with EDB PG AI:

• Run inside your infrastructure
• Fully auditable and version-controlled
• Behavior and API integrations are transparent and governed
• No reliance on third-party "black box" plugins or agents
• Integrated observability and monitoring via Hybrid Manager

Structures enable you to build intelligent, integrated, compliant AI workflows — fully under your control.

Summary

Structures make your AI Factory extensible and production-grade:

• Implement custom business logic
• Integrate with enterprise APIs and systems
• Transform and enrich data pipelines
• Extend Assistants with powerful Tools
• Enable Sovereign AI with full auditability and control

Without Structures, your AI Factory applications cannot fully reflect your business workflows or compliance requirements.

Next steps

• Create a Structure
• Manage Structures in Gen AI Builder
• Create an Assistant
• Rulesets Explained
• Retrievers Explained
• Knowledge Bases Explained
• AI Factory Concepts
• Hybrid Manager: Using Gen AI Builder

Structures turn EDB PG AI from a "chat system" into a full-fledged intelligent application platform — enabling complex, governed AI workflows across your business.