Rioc

Webcam streaming over HTTP and an Ollama vision test (Rioc persona).

Prerequisites

• Python 3.x
• Camera (default index 0)
• For vision_test: local Ollama with llama3.2-vision (e.g. ollama run llama3.2-vision)

Install

From the rioc/ directory:

pip install -r requirements.txt

Run

Webcam stream

Stream the webcam at 640x640, 70% JPEG (MJPEG) for local or cloud consumption:

uvicorn webcam_stream:app --host 0.0.0.0 --port 8000

Then open http://localhost:8000 or http://localhost:8000/video in a browser.

Optional: Local Visual Audit — Run the same Visual Audit (tactical analyst, MiniCPM-V 2.6) in the same process:

1. Have Ollama running with the model: ollama run openbmb/minicpm-v2.6 (once).
2. Enable the audit and start the stream:

ENABLE_LOCAL_AUDIT=1 uvicorn webcam_stream:app --host 0.0.0.0 --port 8000

Audit results print in the server console every few seconds. Optional env: OLLAMA_URL (default http://localhost:11434), OLLAMA_VISION_MODEL (default openbmb/minicpm-v2.6), AUDIT_INTERVAL_SEC (default 5.0).

Vision test (Rioc)

Capture frames and send them to Ollama with the Rioc persona:

python vision_test.py

Note: Both apps use the same camera (index 0). Do not run the webcam stream and vision test at the same time on one machine if they share the device, or the camera may be unavailable to one of them.

Cloud Brain (Visual Audit on AWS)

cloud_brain.py is intended to run on AWS. It pulls frames from the Mac’s MJPEG stream and sends them to vLLM (MiniCPM-V-2_6) via an OpenAI-compatible API for a “Visual Audit.”

Requirements:

1. vLLM running on AWS serving openbmb/MiniCPM-V-2_6 (e.g. vllm serve openbmb/MiniCPM-V-2_6 with the appropriate vision/chat template).
2. The Mac stream reachable at http://<MY_MAC_IP>:8000/video (run the webcam stream on the Mac with uvicorn webcam_stream:app --host 0.0.0.0 --port 8000 and ensure the AWS instance can reach the Mac’s IP).

Environment variables:

Variable	Description	Default
MAC_IP	Mac’s IP (stream at http://<MAC_IP>:8000/video)	localhost
STREAM_URL	Full stream base URL; overrides MAC_IP if set	http://<MAC_IP>:8000
VLLM_BASE_URL	vLLM OpenAI-compatible API base (e.g. http://<vllm-host>:8000/v1)	http://localhost:8000/v1
AUDIT_INTERVAL_SEC	Seconds between audits	5.0
OPENAI_API_KEY	API key for vLLM (use EMPTY if none)	EMPTY

Run:

# On AWS (Mac IP and vLLM host set for your environment)
export MAC_IP=192.168.1.100
export VLLM_BASE_URL=http://vllm-host:8000/v1
python cloud_brain.py