Vercel's Chat SDK now supports Dial via a new vendor-official adapter, enabling bots to send and receive SMS, MMS, and iMessage on real phone numbers. The adapter also handles inbound voice-call transcripts. Each phone conversation maps to a Chat SDK thread, and replies use the standard thread and message APIs. HMAC-verified webhooks and stable per-conversation threading are included out of the box.
Nguồn: https://vercel.com/changelog/chat-sdk-adds-dial-support. 8sync News chỉ tóm tắt và dẫn link; bản quyền nội dung thuộc tác giả và nguồn gốc.
Vercel's Chat SDK now supports Photon, enabling developers to build bots that send and receive iMessages in direct and group chats. The integration supports media sharing, native tapback reactions, and HMAC-verified webhooks. Bots can run via Spectrum Cloud, a self-hosted server, or directly on a Mac. Each Photon conversation maps to a Chat SDK thread, and the adapter follows the same handler/subscription patterns as other Chat SDK adapters.
A walkthrough of building a multi-agent support chatbot for react-admin using the OpenAI Agents SDK. The architecture includes a TriageAgent that classifies incoming questions and hands off to either a DocsAgent (which searches a vector store of 300+ markdown docs using fileSearchTool) or a SupportAgent for billing/subscription queries. Input and output guardrails are implemented to block prompt injections and prevent information leaks, including OpenAI's omni-moderation model for harmful content filtering. The bot is exposed via a Hono API with a custom React frontend, streaming responses to the UI. The post also touches on tracing, human-in-the-loop, and voice agent features available in the SDK.
Meta ran a covert project codenamed 'Cannes' in which hundreds of contractors, managed by a firm called Covalen, created fake under-18 accounts to probe the safety systems of ChatGPT, Gemini, and Character.AI. Over 45,000 prompts were sent in a single round, many involving suicide, self-harm, eating disorders, and sexual content — some adopting the voice of a child in crisis. None of the targeted companies authorized the testing, and all three bar such activity in their terms of service. Meta defends the work as standard industry benchmarking, but AI safety experts call it a 'governance gray zone' that blurs safety evaluation with competitive intelligence gathering. The revelations land as the FTC is already investigating AI and child safety, and EU regulators are watching under the AI Act and Digital Services Act.