Overview

Nyra Agent represents an advanced conversational AI entity, seamlessly integrating Nyra’s core capabilities with cutting-edge technologies such as Gemini 2.0 Live, OpenAI Realtime, and RTC protocols. This architecture empowers the agent with real-time sensory and communicative functionalities, including dynamic vision, auditory processing, and speech synthesis. Furthermore, Nyra Agent ensures robust interoperability with leading workflow ecosystems, such as Dify and Coze, providing unparalleled versatility and adaptability in diverse operational contexts.

Links

  • NYRA Framework

Architecture

The Nyra Agent project is organized into the following major components, offering clarity and extensibility for developers:

  1. Agents Module: Hosts the foundational logic, executable binaries, and example implementations essential for constructing and deploying AI agents. Within the Agents directory, a dedicated nyra_packages,subfolder provides a comprehensive collection of prebuilt, extensible components. These modular extensions enable developers to efficiently design, customize, and optimize agents for specialized tasks and workflow integrations.

  2. Dev Server: Manages backend operations, orchestrates agent workflows, and oversees extension interactions.

  3. Frontend Server: Operates on port 8080, serving the user interface, processing HTTP requests, and delivering static resources.

  4. Extensions Module: Provides modular components for seamless integration with LLMs, TTS/STT systems, and external APIs, facilitating extensive customization options.

  5. Interactive Playground: A configurable sandbox environment for testing, tuning, and refining agent behaviours.

  6. Demonstration Environment: A fully deployable setup showcasing practical implementations and real-world use cases of Nyra Agent.

Dockers Container

There are two Docker containers in TEN Agent:

  • nyra_agent_dev: The primary development container driving Nyra Agent encompasses the core runtime environment, essential development tools, and all necessary dependencies for building and running agents. This container enables streamlined workflows, allowing you to execute commands like task use to construct agents and task run to launch the web server.

  • nyra_agent_playground : A specialized container running on port 3000, dedicated to the web frontend interface. It delivers precompiled frontend assets and offers an interactive platform for configuring modules, selecting extensions, and testing agents. The Playground UI provides intuitive tools for visually selecting graph types (e.g., Voice Agent, Realtime Agent), customizing modules, and managing API configurations, streamlining the agent development process.

  • nyra_agent_demo : A deployment-oriented container running on port 3002, designed to showcase a production-ready example setup. It illustrates how users can deploy their configured agents in real-world scenarios, with all essential components bundled for seamless deployment and operation.

Agents

The Agents folder is the core of the project, cotaining:

  • Fundamental binaries and reference implementations that define and demonstrate agent behavior.

  • Dynamic resources enabling adaptable configurations for diverse AI applications and workflows.

  • A suite of utilities empowering developers to design, customize, and optimize AI agents effectively.

The structured design of the Agents folder empowers developers to create agents customized for specific applications, such as voice assistants, chatbots, or task automation workflows.

Demo

The Demo folder provides a deployment-ready environment for showcasing Nyra Agent live. It includes:

  • Sample setups optimized for deploying agents in live environments.

  • Ready-to-use examples showcasing the framework's functionality and potential.

  • Utilities for presenting real-world applications to users, clients, and collaborators effectively.

Playground

Once the playground is operational, users can leverage the module picker to:

  • Choose and customize extensions from a library of prebuilt, modular components.

  • Experiment with diverse AI models, TTS/STT engines, and real-time communication frameworks.

  • Safely validate and refine agent behaviors within a controlled, sandboxed environment.

The playground acts as a central innovation platform, enabling developers to seamlessly experiment with and optimize their AI systems.

PreviousScenariosNextGetting Started

Last updated