Changelogs: Week of July 6, 2026

This week, Washington DC joined the PolarGrid network, vision/image input shipped in the SDKs with server-side rollout underway, and voice agents are now live in production.

Washington DC is live, PolarGrid now spans 11 cities

The new DC edge node brings PolarGrid to 11 cities across North America. For teams on the US East Coast, Washington joins the network as a low-latency option for workloads that were previously routing through New York or further.

The node is immediately available as a selectable region in the dashboard, playground, and both the JS and Python SDKs. Intelligent auto-routing, health monitoring, and automatic failover are active from day one. If you're using region auto-select, your traffic will be routed through the DC automatically when it's the closest healthy node.

Vision/image input

Both the JS and Python SDKs now accept image input in chat completions using the same OpenAI-compatible format; pass an image URL or base64-encoded content in the messages array. No new API shape to learn, no new integration path.

Server-side vision support is rolling out across the fleet, starting on staging. Full regional availability is on the way.

Voice agents are in production

The full real-time voice stack: speech-to-text → LLM → speech, is live and serving production traffic. The pipeline handles the complete turn end-to-end: a caller speaks, audio hits PolarGrid's edge, the model responds, and synthesized speech is returned, all within a single latency budget.

For teams building voice-based applications, this is the infrastructure layer. You bring the use case; PolarGrid handles the real-time stack. The same sub-400ms pipeline that runs our own pilots is available for your deployments today.