# GetStream/Vision-Agents

Open Vision Agents by Stream. Build voice and vision agents quickly with any model or video provider. Uses Stream's edge network for ultra-low latency.

Repository: https://github.com/GetStream/Vision-Agents
Canonical: https://ross.abutalabs.com/products/vision-agents
Homepage: https://visionagents.ai
Language: Python
License: Apache-2.0
License Family: permissive
Topics: ai, ai-agents, vision-ai, voice-ai, video-agents, agentic-ai, agents, realtime, stt, tts, video-ai
Last push: 2026-08-26T18:22:58+00:00

## Health v2 (maintenance only)
Score: 84/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 97, longevity 27
- inputs: {"age_days": 387, "days_push": 7, "days_rel": 20, "gap_med": 4.0, "n_releases_24m": 59}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 8100, forks 680 (observed 2026-08-28T04:10:12.695529+00:00)

## What it is
An open-source Python framework by Stream for building low-latency real-time voice and video AI agents. It provides 35+ provider plugins (OpenAI, Gemini, Deepgram, ElevenLabs, YOLO, Roboflow) and uses Stream's WebRTC edge network for sub-500ms join and sub-30ms audio/video latency.

## Use cases
- build a real-time voice agent that talks to users in the browser
- create an AI sports or golf coach using YOLO pose detection and Gemini
- build a smart security camera with face recognition and package detection
- create a phone support agent that answers inbound Twilio calls with RAG
- build a live sports commentator that tracks players and generates play-by-play
- create an interactive lip-synced avatar that sees and hears users
- run object detection models on live video frames in real time

## When to choose
- you need real-time multimodal agents that process live audio and video with low latency
- you want to mix vision models like YOLO or Roboflow with LLMs like Gemini or OpenAI in one pipeline
- you want provider-agnostic STT, TTS, LLM, and vision plugins with a consistent interface
- you need client SDKs across React, iOS, Android, Flutter, React Native, or Unity
- you want to build telehealth, live coaching, or voice support agents quickly

## When to avoid
- you only need offline or batch video analysis with no real-time interaction
- you want to avoid Stream's edge network or paid services and have no WebRTC infrastructure of your own
- you need a simple text-only chatbot without voice or video
- you require a non-Python backend language for your agent logic

## Facets
- artifact type: framework
- maturity: active
- function: agent-framework, speech-recognition, tts, computer-vision, llm-inference, rag, chatbot, sdk
- domain: artificial-intelligence, computer-vision, speech-processing, large-language-models
- platform: python, cross-platform, cloud
- tags: voice-ai, video-ai, realtime-agents, webrtc, stt, tts, multimodal, stream-video, yolo, telehealth, ai-agents, video, real-time, web-server, docker

## Member repositories
- GetStream/Vision-Agents (main) score 84

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:12.695529+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:30:42.466799+00:00, confidence not recorded.
  - readme: https://github.com/GetStream/Vision-Agents (fetched 2026-08-28T04:10:12.695529+00:00, sha a26f242a36f6)
  - homepage: https://visionagents.ai (fetched 2026-08-29T08:28:48.713206+00:00, sha 499f0e4b66f8)
  - site_page: https://visionagents.ai/introduction/quickstart (fetched 2026-08-29T08:28:48.724036+00:00, sha 2146b53149f6)
  - registry_pypi: https://pypi.org/pypi/vision-agents/json (fetched 2026-08-29T08:28:48.725960+00:00, sha 12f26363d708)
  - site_page: https://visionagents.ai/integrations/introduction-to-integrations (fetched 2026-08-29T08:28:48.722031+00:00, sha 9d58993b2abc)
- Data as of 2026-08-30T08:39:29.467469+00:00.