Async AI pipelines for Rust - Build intelligent, fault-tolerant data processing flows using the Tokio runtime

AgentFlow is a high-performance framework for building AI pipelines in Rust. Inspired by pipecat-ai, it provides a robust async architecture for creating complex data processing workflows with built-in fault tolerance and real-time capabilities. The core is powered by Tokio for scalable, concurrent execution.

• Async Architecture: Built on Tokio for reliable, concurrent task execution
• Type-Safe Pipelines: Strongly-typed message passing ensures correctness at compile time
• Fault Tolerance: Supervisor trees automatically restart failed tasks without disrupting the entire pipeline
• Real-Time Processing: Optimized for low-latency audio, video, and text processing
• Extensible Processors: Rich ecosystem of pre-built processors for common AI tasks
• Hot Reloading: Update pipeline logic without stopping the system
• Observability: Built-in monitoring, metrics, and debugging tools

AgentFlow organizes processing into three core concepts:

• Transports: Handle input/output streams (audio, video, text)
• Processors: Transform and process data streams
• Tasks: Manage background processing and coordination

• Voice Assistants: STT → LLM → TTS pipelines with real-time audio processing
• Video Analysis: Computer vision → LLM reasoning → action generation
• Document Processing: OCR → text analysis → structured output generation
• Real-time Translation: Speech → transcription → translation → synthesis
• Multi-modal AI: Combining vision, language, and audio processing processors

Early Development - Core architecture being built. Not yet ready for production use.

Built with ❤️ and ⚡ for the Rust AI community