Vision-Agents and vision-agent
Stream's Vision-Agents framework (A) is the underlying open-source platform that the Vision Possible Hackathon project (B) was built upon, making them ecosystem siblings where B represents a reference implementation or downstream application of A's architecture.
About Vision-Agents
GetStream/Vision-Agents
Open Vision Agents by Stream. Build Vision Agents quickly with any model or video provider. Uses Stream's edge network for ultra-low latency.
Features a pluggable processor pipeline for computer vision models (YOLO, Roboflow, PyTorch/ONNX) that run before LLM calls, plus native integrations with OpenAI Realtime, Gemini Live, and Claude for streaming AI responses. Includes voice integration via Deepgram/AssemblyAI for STT and ElevenLabs/Cartesia for TTS, turn detection with VAD, tool calling via MCP, and production-ready HTTP server with Prometheus metrics for horizontal scaling and Kubernetes deployment.
About vision-agent
rupac4530-creator/vision-agent
Production-grade multi-modal AI platform — 17 real-time vision & audio tabs, 22 SDK modules, 7-tier LLM cascade, 37+ endpoints. Built for Vision Possible Hackathon by WeMakeDevs x Stream.
Related comparisons
Scores updated daily from GitHub, PyPI, and npm data. How scores work