Agent Glasses Memory Core

Open-source demo backend for an AI smart-glasses memory pipeline.

This project demonstrates how smart glasses, a mobile app, or any wearable device could send captured events into an agentic backend.

The backend receives events, queues them, processes them through simple demo agents, stores structured memory, and exposes the result over an API.

What this is

A technical demo showing:

• smart-glasses event ingestion
• voice/text capture flow
• Redis-based async queue
• MySQL persistence
• agent orchestration
• structured memory output
• simulated notification back to glasses/mobile
• clean GitHub-ready project structure

What this is not

This repository is intentionally limited.

It does not include:

• real Meta / Ray-Ban API integration
• real speech-to-text provider
• real Gmail / Calendar integration
• production authentication
• billing
• customer portal
• production prompts
• commercial field-service logic
• real notification delivery
• mobile app
• SaaS deployment layer

This is a public demo core, not a full commercial product.

Concept

Smart Glasses / Mobile App
        |
        v
POST /api/v1/glasses/events
        |
        v
MySQL: raw captured event
        |
        v
Redis queue
        |
        v
Agent Worker
        |
        +--> CaptureAgent
        +--> MemoryAgent
        +--> InsightAgent
        +--> NotificationAgent
        |
        v
MySQL: structured memory + agent logs

Why this exists

Smart glasses are only an input/output device.

The real product is the backend layer:

• capture
• classify
• remember
• retrieve
• notify
• act

This demo shows the architecture without exposing real commercial workflows.

Quick start

1. Create project

chmod +x create_agent_glasses_memory_core.sh
./create_agent_glasses_memory_core.sh
cd agent-glasses-memory-core

2. Start MySQL and Redis

docker compose up -d mysql redis

3. Activate Python environment

source .venv/bin/activate

4. Run API

./scripts/run_api.sh

5. Run worker in second terminal

source .venv/bin/activate
./scripts/run_worker.sh

6. Send demo smart-glasses event

./scripts/demo_glasses_event.sh

Endpoints

Health

GET /health

Create glasses event

POST /api/v1/glasses/events

Example:

{
  "device_id": "glasses-demo-001",
  "event_type": "voice_note",
  "transcript": "Remember this idea. Smart glasses are not the product. The backend memory layer is the product.",
  "context": {
    "source": "demo",
    "location_label": "office",
    "confidence": 0.92
  }
}

List captured events

GET /api/v1/glasses/events

List structured memories

GET /api/v1/memories

List agent logs

GET /api/v1/agent-logs

Example output

{
  "memory_type": "idea",
  "summary": "Smart glasses are an interface. The backend memory layer is the actual product.",
  "tags": ["demo", "smart-glasses", "memory"],
  "suggested_action": "Save as product insight",
  "notification_preview": "Captured idea: backend memory layer is the product."
}

Commercial boundary

The public version demonstrates the architecture only.

A commercial product would usually add:

• hosted environment
• authenticated users
• real wearable/mobile app
• real STT
• real integrations
• workflow-specific agents
• domain templates
• approvals
• analytics
• billing
• team management
• compliance/audit features

License

This demo includes a placeholder license. Replace it with AGPL-3.0, Apache-2.0, or a dual-license model depending on your commercial strategy.