Now in public beta

Visual Memory
for AI

Store and search images using semantic embeddings. Give your AI agents the ability to remember what they see.

Quick Start

quickstart.py

Install with: pip install erised

Everything you need for visual AI

Erised uses state-of-the-art ColQwen2.5 embeddings to understand images semantically, not just by pixels.

Semantic Understanding

Search images by meaning, not keywords. Find 'code editor' and get results showing IDEs, terminals, and coding sessions.

User Isolation

Each user's memories are isolated. Filter searches by user_id for multi-tenant applications.

Simple API

RESTful API with Python SDK. Add images, search by text, list and delete memories.

Use Cases

Built for real applications

See how Erised powers visual memory across different use cases.

Computer Use Agents

Give AI agents visual memory across sessions

Store screenshots during browsing sessions. Your agent can recall what it saw hours or days ago using semantic search.

Multi-session contextSemantic screenshot searchReduce token costs
example.py

How it works

01

Add Images

Upload screenshots, photos, or any images. Each image is processed to generate semantic embeddings.

02

Search by Text

Query your memories using natural language. 'Show me the settings page' or 'Find the error message' just work.

03

Get Results

Receive ranked results with similarity scores. Use in your AI agents, apps, or workflows.

Ready to get started?

Give your AI the gift of visual memory. Start building in minutes.