Back to browse
GitHub Repository

AI-powered screen memory — captures, analyzes, and lets you search/chat your screen history. Powered by Gemma 4 . 100% local, 100% private.

123 starsPython

Running a vision model on every screenshot on-device

by skye0110·Jun 23, 2026·1 point·0 comments

AI Analysis

●●●BangerBig BrainZero to One

Privacy-first Recall alternative running Gemma 4 vision entirely on-device.

Strengths
  • Content-change detection captures only when screen actually changes, not fixed intervals.
  • Hybrid search combines MiniLM embeddings with FTS5 for semantic and keyword results.
  • MCP server integration works with Claude, Cursor, and VSCode out of the box.
Weaknesses
  • AI memory space getting crowded with Rewind, Recall clones emerging.
  • Gemma 4 vision models require significant local compute resources.
Category
Target Audience

Privacy-conscious developers and knowledge workers

Similar To

Rewind.ai · Microsoft Recall · ActivityWatch

Similar Projects

AI/ML●●Solid

I run a vision model on every screenshot, locally, on a 4GB GPU

Runs multimodal screen memory on 4GB VRAM while Microsoft Recall requires high-end NPU.

Dark HorseSolve My Problem
skye0110
36710d ago
AI/ML●●●Banger

Qwen 3.5 running on a $300 Android phone – on-device, open source

Comprehensive offline AI suite (text, vision, images) on $300 phones—genuinely complete.

Zero to OneWizardry
ali_chherawalla
6103mo ago