fix(talk): steering an active voice agent consult fails with PostgreSQL agent stores (#189036) * fix(talk): steer active voice consults with PostgreSQL agent stores Steering, follow-up and cancel of an active Talk run built the caller's tool authority overlay with a synchronous agent-database read on the Gateway thread, which PostgreSQL agent stores refuse. Add an optional getToolAuthorityOverlayAsync to controlRealtimeVoiceAgentRun, awaited at selection and in each injection admission's preparation, and read the session entry through getSessionEntryAsync for every Talk steering surface. The synchronous compatibility check validates the overlay prepared for that admission. The synchronous SDK callback is unchanged. * test(talk): pin async steering authority at the final effect Cover the asynchronous caller-authority path of controlRealtimeVoiceAgentRun against the real session controller: an allowed caller is queued, a wrong caller is rejected by the run's tool authority, and revocation or reassignment during the injection's own authority read never reaches backend enqueue. * refactor(talk): keep the async steering authority read off the public SDK The public realtime-voice controlRealtimeVoiceAgentRun keeps its released synchronous callback. Gateway Talk callers use the internal controlTalkAgentRun, which reads caller policy through the worker-backed session reader. * test(talk): state why the handler suite isolates run control
fix(talk): steering an active voice agent consult fails with PostgreS…
Find Your Interest
- GPU giants throw HPC users some tasty crumbs
- fix: keep automations running when someone renames a session group or…
- How NVIDIA OpenShell Reference Architecture Secures and Governs AI Agent Workloads at Scale
- Start and Restore GPU Workloads in Seconds with NVIDIA Snapshot
- Keep Distributed Training Running with Automated Workload Recovery
- Making Kubernetes AI-Ready for NVIDIA Vera Rubin on Day One
- Memory-Enforced GPU Sharing in Kubernetes with NVIDIA CUDA 13.4
- Nvidia just started a new chip war
- NVIDIA's H100: The Powerhouse GPU Revolutionizing Deep Learning
- Accelerating Large Language Models: The H100 GPU’s Role in Advanced AI Development





