Features#

Orkes is designed with a focus on providing a clear, flexible, and powerful framework for building complex workflows. Here are some of the key features that make this possible:

Graph-based Architecture#

At the core of Orkes is the OrkesGraph, a powerful abstraction for defining and managing computational workflows. This graph-based architecture allows you to represent complex logic as a collection of nodes and edges, providing a clear and intuitive way to visualize and control the flow of execution.

The interface for defining these graphs is inspired by the well-known NetworkX library, providing a familiar and powerful paradigm for those with experience in graph-based programming.

Key features of the graph architecture include:

  • Stateful Execution: A shared state object is passed between nodes, allowing for seamless data flow and management throughout the graph’s execution.

  • Flexible Control Flow: Orkes supports both simple, unconditional edges for linear workflows and powerful conditional edges for creating branching logic and loops. This allows you to build dynamic and adaptive systems.

  • Simple, Pythonic Nodes: Nodes in an Orkes graph are just plain Python functions, making it easy to integrate your existing code and logic into a graph-based workflow.

Graph Traceability and Visualization#

Understanding and debugging complex AI workflows is a major challenge. Orkes addresses this with a built-in traceability and visualization system. When you run a graph, Orkes can generate a detailed execution trace that captures the state, timing, and data flow at every step.

This trace can then be used to generate a self-contained, interactive HTML file that provides a visual representation of the graph’s execution. This allows you to:

  • Visualize the execution path: See exactly which nodes were executed and in what order.

  • Inspect the state at any point: Understand how the data in your graph is being transformed.

  • Debug with ease: Quickly identify bottlenecks, errors, and unexpected behavior.

Pluggable LLM Integrations#

Orkes provides a flexible and extensible system for integrating with Large Language Models (LLMs). The UniversalLLMClient provides a unified interface for interacting with different LLM providers, and the LLMFactory makes it easy to configure and switch between them.

Out-of-the-box, Orkes includes support for:

  • OpenAI-compatible APIs

  • Anthropic’s Claude

  • Google’s Gemini

You can also easily implement your own custom strategies to connect to any other LLM service or local model.

Agent and Tool Support#

While the graph-based architecture is the core of Orkes, it also provides support for building and using agents and tools. You can define custom tools using the OrkesToolSchema and use them within your graph’s nodes to interact with external APIs, databases, or other services. This allows you to build sophisticated, interactive systems within the structured and controllable environment of an Orkes graph.