Structures

Structures are the building blocks of advanced AI workflows in AI Factory. They allow you to orchestrate multi-step pipelines, integrate external APIs and systems, run data transformations, and extend Assistants with custom Tools and logic — all under your governance.

Structures make it possible to go beyond basic chatbots and build enterprise-grade AI applications that combine reasoning, retrieval, action, and compliance.


Key features of Structures

  • Based on Griptape Python agents, pipelines, workflows
  • Deployed from:
  • Data Lake zip file
  • GitHub repository
  • Provided samples
  • Configurable via:
  • Environment variables
  • Structure Config files (YAML / Python)
  • Executable via:
  • PG.AI Web Console
  • API
  • Third-party integrations

At runtime, Structures can run:

  • On demand via UI
  • Programmatically via API
  • Triggered by Assistant Tools
  • Embedded in Data Source transformation pipelines

When to use Structures

Use Structures when:

  • You need multi-step AI reasoning and workflows
  • You want to call external APIs from an Assistant or a pipeline
  • You need to transform data at ingest time or pre-indexing
  • You want to implement custom retrieval logic beyond basic RAG
  • You want to build domain-specific AI workflows that reflect your business logic

Structures power agentic behavior, complex pipelines, and production use cases where advanced control is required.


Concepts and Explained


How-To Guides


Example Use Cases

  • Document transformation pipeline Extract, enrich, and reformat documents for downstream Knowledge Base indexing.

  • Custom API-driven Assistant Tool Expose a Tool that calls your internal pricing system or CRM.

  • Complex summarization agent Build a Structure that chains summarization, translation, and tagging.

  • Transaction categorization pipeline Run a multi-step process to classify and tag financial transactions.

  • Dynamic retrieval agent Implement a Retriever that combines semantic and rule-based logic.


Getting started

See Create a Structure for a full step-by-step guide.

Typical workflow:

  1. Package your Griptape Structure as a zip file.
  2. Upload to your Data Lake.
  3. Create a Structure in Gen AI Builder.
  4. Configure Structure parameters.
  5. Execute Structure:
  • Directly
  • As part of Data Source transformation
  • As an Assistant Tool



AI Factory Structures give you the power to turn AI into real applications — secure, governable, and extensible across your enterprise.

Start building today.



Could this page be better? Report a problem or suggest an addition!