September 24, 2026
Mastra's update introduces observability capability negotiation, eager tool execution during streaming, trace aggregation planning, and new integrations for Discord, sandbox credential materialization, and MongoDB automated embeddings.
原始内容为英文；当前页面提供中文导航与来源说明，具体事实请以原文为准。
为什么值得关注
Mastra has released core version 1.71.0 along with ecosystem additions. Observability stores can now negotiate features with servers and clients using getFeatures() and GET /observability/capabilities, preventing 500 errors on unsupported discovery calls. A new planTraceAggregate() method validates aggregateTraces() queries into a TrustedTraceAggregatePlan for storage backends. Multi-tool streaming runs now support eager tool execution, initiating tool execution as soon as individual tool arguments complete rather than awaiting full step streaming completion (with exceptions for tools requiring approval, background execution, or suspension). New integrations include @mastra/discord, sandbox credential materialization via @mastra/connect environment(), and MongoDBVector automated embeddings. Additionally, @mastra/playground-ui introduces breaking changes to the TaskList API by removing title, TaskListHeader, and hideWhenEmpty.
可执行摘要
Mastra core 1.71.0 introduces observabilityStorageCapabilities negotiation, eager tool execution during multi-tool streaming steps (default on, opt-out via eagerToolExecution: false), planTraceAggregate() for storage backends, and breaking changes in @mastra/playground-ui TaskList API.
- Agent 实用度
- 82/100
- 可信度
- 96%
- 机器格式
- JSON + Markdown
开发者应核对什么
- Update @mastra/playground-ui usage by removing references to title, TaskListHeader, and hideWhenEmpty from TaskList implementations.
- Evaluate streaming workflows with multi-tool calls; if dependent on previous behavior where execution waited for full stream completion, configure eagerToolExecution: false.
- Implement getFeatures() on custom ObservabilityStorage classes to expose supported discovery, metric, and feedback features to clients and Studio.
- Review backend implementations of aggregateTraces() to utilize planTraceAggregate() for consistent validation and handling.