TEN-framework/ten-framework is an open-source, distributed, cloud-edge collaborative AI agent framework designed for real-time multimodal interactions. It supports multiple programming languages (C++, Go, Python, JS/TS) and platforms (Windows, Mac, Linux, mobile), enabling developers to build complex AI applications by integrating various extensions, audio-visual tools, databases, and monitoring systems. The framework aims to provide high performance, low latency, and flexibility for both edge and cloud deployments, allowing for cost and performance optimization.

How It Works

TEN operates as a core application managing extensions and data flow based on graph configurations. It supports a modular, drag-and-drop approach to agent construction, facilitating the integration of AI with diverse tools and services. The framework emphasizes real-time agent state management for dynamic responsiveness and allows for edge-cloud integration, leveraging local compute for privacy-sensitive tasks and cloud models for enhanced performance.

Quick Start & Requirements

• Install/Run: Primarily uses Docker Compose for local development.
• Prerequisites:
  • Docker / Docker Compose
  • Node.js (LTS) v18
  • API Keys: Agora App ID/Certificate, OpenAI API key, Deepgram ASR, Elevenlabs TTS.
  • Minimum System: 2 Core CPU, 4 GB RAM.
• Setup: Building the default agent graph takes approximately 5-8 minutes.
• Links: Documentation, Playground Guide, Demo Guide

Highlighted Details

• Supports real-time multimodal interactions with low latency.
• Enables edge-cloud integration for optimized cost and performance.
• Showcases integrations with various LLMs (Llama 4, Gemini, DeepSeek) and hardware (ESP32).
• Offers a Playground for agent customization and a Demo server for deployment.

Maintenance & Community

• Active community with Discord server (https://discord.gg/VnPftUzAMJ) and X presence.
• Open to contributions with clear guidelines.

Licensing & Compatibility

• License: Apache License, Version 2.0, with additional restrictions noted in the LICENSE file.
• Compatibility: Generally compatible with commercial use, but specific restrictions should be reviewed in the LICENSE file.

Limitations & Caveats

The framework is described as having an alpha version of its TMAN Designer. Some cloud service deployment options are listed as "coming soon."