Engineering
How to Build a Social Media App: A Technical Guide
Building a social media app means a single user action must propagate to potentially millions of other users in real time, while staying fast, safe, and cheap. Every feature touches every other feature. And the hard problems shift as you scale. At 100K users, it's the database. At 1M users, it's the fan-out strategies. At
5 Lessons from Building a Voice AI Prescription Agent
What we learned building a prescription-intake voice agent with Stream's Vision Agents SDK and AssemblyAI: five lessons for safety-critical AI agents.
Real-Time Voice AI For Phone Support In 2026
A phone support voice AI needs to react in real time - start replying within 300ms, stop the moment a caller interrupts, and handle the telephony edge before the model sees the audio. This guide covers the full pipeline - streaming STT, TTS, turn detection, barge-in, and the latency budget that keeps calls feeling natural.
Warm Transfer vs Cold Transfer: Designing AI-to-Human Escalation in Real-Time Voice Systems
Warm transfer carries the conversation's context to the human agent - the summary, the intent, the caller's mood - so the human can pick up mid-story. Cold transfer carries only the audio, and the caller starts over. Building warm transfer in a real-time voice system means solving three problems when to escalate, how to generate structured context in parallel with the live call, and how to hand off the audio session without breaking it.
What Is a Vision Agent? Real-Time AI That Can See and Hear
A vision agent watches a live video and audio stream and acts while things are still happening. Here is what a vision agent is, how it differs from batch vision pipelines and voice agents, and how to build one in about 18 lines with Vision Agents, the open-source Python framework from Stream.
The End of the Orb: Building AI Agents That Feel Present
TLDR: Agents these days are blind and not very engaging, so we decided to team up with Anam and Inworld to build an agent using Vision Agents that feels personal and aware of the world around you. Give it a try here. Most voice agents today are blind. They hear words, convert them to text,
Chat Application Architecture, Explained
TLDR; Wide-column stores like Cassandra handle messages while Redis holds read state, because each subsystem's access patterns differ significantly. Presence alone generates a write on every connect, disconnect, and heartbeat, making it orders of magnitude more write-heavy than messages. End-to-end encryption prevents the server from searching, moderating, or generating push notification previews, which transport-plus-at-rest encryption
Scaling Event-Driven Systems Without Compromising Mobile App Stability
Event-driven architecture is nothing new. IBM MQ shipped in 1993. JMS has been around since 1998. Kafka launched in 2011. But for most of that history, event-driven patterns were for specialized domains. Most developers never touched them. That's changed. Real-time mobile features, such as chat, activity feeds, live collaboration, or presence indicators, have pushed event-driven
Vision Agents v0.5.0 Release: Local Hardware I/O, Anam Avatars, and Faster Deepgram TTS
It's been a busy period since our last release, and now it's time to share Vision Agents v0.5.0: a step toward making production-grade multimodal AI agents easy to build and deploy. While previous versions laid the groundwork for real-time voice, video, and Vision Agents, v0.5.0 focuses on stability at scale and even more expressive