Enforced unique IDs across Agents, Teams, and Workflows to prevent collisions

Finer-grained Milvus queries for faster, more accurate retrieval

Milvus search and async_search now support radius, range_filter, and async search_parameters. These controls help teams tune recall vs. precision and reduce tail latency in high-throughput workloads.

Standardized A2A endpoints improve interoperability and client simplicity

We introduced conventional A2A endpoints — including Agent Card retrieval — and aligned run endpoints and payloads to the updated protocol.

Stream reasoning in real time to speed iteration and oversight

You can now stream reasoning chunks whenever a reasoning model is used. A new ReasoningManager coordinates streaming and lifecycle, giving teams earlier visibility into model thinking, faster…