The Agent Architect #46: A2A Protocol (Agent2Agent)

Pattern of the week

A2A Protocol (Agent2Agent)

What:

Agents discover each other's capabilities via standardized cards, then exchange messages with task IDs to coordinate work across platforms.

When to use it:

Multiple independent agents from different systems need to collaborate on long-running tasks with real-time state synchronization.

Watch out:

High implementation complexity and operational overhead often outweigh benefits unless you genuinely need cross-platform interoperability at scale.

This week in agentic AI

• vLLM Delivers Day-0 Support for Kimi K3.1 ProvLLM

  vLLM supports Kimi K3.1 Pro with sparse attention, packed KV cache, and fused low-latency decode.

• Speed Up Image Generation with FLUX 3 Image Klein on NVIDIA Blackwell GPUsNVIDIA Developer

  FLUX 3 Image Klein achieves faster image generation on NVIDIA Blackwell GPUs.

• CodeQuant2: Enhanced Outlier Mitigation and Adaptive Clustering for Low-Precision LRMsarXiv cs.AI

  Large reasoning models show larger activation fluctuations in early decoding steps, affecting low-precision quantization.

• BitDecoding-FP4: Native FP4 KV Cache Decoding via Algorithm-System Co-DesignarXiv cs.AI

  BitDecoding-FP4 enables native FP4 key-value cache decoding for long-context LLM serving through algorithm-system co-design.

• A Closer Look at Phonon TTS. Launching High-Quality On-Device Speech That Tops the Benchmarks.Hugging Face

  Phonon TTS is an on-device text-to-speech system that achieves high quality and tops benchmark performance metrics.