Report Abuse

Basic Information

TEN-framework is an open-source ecosystem and codebase for creating, customizing, testing, and deploying real-time multimodal conversational AI agents. The repository provides a complete agent showcase and developer tooling including TEN Agent, TMAN Designer (a low/no-code workflow UI), TEN VAD, TEN Turn Detection, StoryTeller image generation, and a portal for documentation. It targets real-time interactions with voice, vision, avatars and screenshare detection and documents integration points for ASR, TTS, LLMs and MCP servers. The README includes setup instructions, examples, a playground, environment and dependency guidance, and deployment options such as Docker and GitHub Codespaces. The project bundles example agents, extension modules, and guides for hardware integration to run on devices like ESP32-S3, enabling developers to iterate locally and prepare services for release builds.

Links

Categorization

App Details

Features
Real-time multimodal capabilities: voice, vision, real-time avatars and screenshare detection. Low/no-code TMAN Designer for building and editing agent workflows, graphs and apps. Ready-to-use extensions and tools such as StoryTeller image generation, Weather Check, Web Search, and integrations with MCP servers and other LLM platforms. Audio pipeline components and integrations: ASR via Deepgram, TTS via ElevenLabs, and low-latency voice activity detection via TEN VAD. Turn detection and full-duplex dialogue support with TEN Turn Detection. Examples, a Playground, and templates in the ai_agents folder to bootstrap agents. Deployment and developer workflows including Docker Compose, container releases, and running in GitHub Codespaces. Documentation, multilingual READMEs, community channels and contribution guidelines.
Use Cases
TEN-framework helps developers and teams prototype and ship interactive conversational agents that combine speech, vision and avatar experiences. It reduces engineering overhead by providing prebuilt modules for voice activity detection, turn handling, ASR/TTS wiring, and LLM integrations, plus a visual designer for non‚Äëcode configuration. The included examples, playground and deployment recipes enable local development, hardware testing on boards like ESP32-S3, and production packaging with Docker. Support for MCP servers and multiple LLM platforms lets projects reuse external models and scale inference. Multilingual docs, community channels and clear contribution guidance make it easier to adopt, extend and operate realtime agents for demos, research, or product integrations.

Please fill the required fields*