v1.0.68 @mastra/core@1.61.0
mastra-ai/mastraAug 24, 2026 → Aug 24, 20260 days between releases
What's New
Adds caller-driven experiments, workflow concurrency safety, and graceful shutdown options.
Breaking Changes
- Concurrent workflow resumes now return HTTP 409 conflicts instead of allowing race conditions
- Graceful shutdown requires new configuration options like drainTimeout
New Features
- Caller-driven experiments for external orchestrators like Temporal
- Graceful shutdown options with HTTP drain controls
- First-class 'while-active' message delivery in sessions
- Workflow resume concurrency safety with atomic claims
- Multi-turn LLM judge scorer for evaluations
Improvements
- Improved shutdown draining behavior and cleanup reliability
- Workflow state updates use an expectedStatus guard
- Caller-driven experiments decouple experiment loops from Mastra
- First-class message delivery preserves accurate transcripts
- Experiments are fully compatible with Studio views
Bug Fixes
- Fixed workflow resume race conditions with atomic claims
- Fixed server shutdown reliability with improved draining
Upgrade Notes
You must configure new shutdown options like drainTimeout and handle potential HTTP 409 errors when resuming workflows concurrently. Existing code relying on legacy workflow resume behavior may need updates to handle the new atomic claim mechanism.
The Two Releases
Meshery v1.0.68
Aug 24, 2026· 2 months ago
Fixes a critical server boot panic that occurred when seeding models under a pinned provider. This release primarily addresses a fatal error preventing the server from starting correctly.
Full release notesAugust 21, 2026
Aug 24, 2026· 2 months ago
Enabled caller-driven experiments for external orchestrators like Temporal and improved workflow resume safety with atomic claims. Added graceful shutdown options and a multi-turn LLM judge scorer.
Full release notes