Cofounder Vision

OpenCV 5 perception for an agent that learns your workflows from one screen recording

Connecting to AWS Lambda…

Learn from a recording

Record yourself doing a task once. OpenCV splits it into steps: what you clicked, where, and what changed.
MP4 or MOV, up to 40 MB. Uploaded straight to a private, encrypted S3 bucket with a 10-minute presigned URL, then deleted right after analysis.

JSON API

The same endpoints the demo uses. Images are base64 in JSON; every response carries an X-Request-Id that matches one structured log line in CloudWatch.
GET/health
OpenCV version, arch, build
POST/api/locate
{screenshot_b64, template_b64, min_score?}
POST/api/verify
{before_b64, after_b64, roi?}
POST/api/upload-url
{content_type, size?} → {url, key}
POST/api/segment
{key, full_targets?} | {video_b64} | {sample}
GET/api/samples
bundled sample inputs
curl -s ORIGIN/health
curl -s -X POST ORIGIN/api/locate -H 'content-type: application/json' \
  -d "{\"screenshot_b64\":\"$(base64 -i screen.png)\",\"template_b64\":\"$(base64 -i button.png)\"}"