How https-deeplearning-ai/sc-on-device-ai-memory works
The Qdrant Edge Memory Robot is an on-device object-memory assistant. Its main workflow detects and tracks camera objects, embeds stable crops for recognition, and lets users teach names by voice; later sightings and spoken questions can retrieve stored memories. The browser interface also supports browsing and managing memories. The graph follows the documented architecture and shown source references; details of unsampled runtime paths and voice-action dispatch are represented conservatively.
Main entry points: live.py Live app · core.py Robot workflow · memory.py Memory operations · server.py Web server
Components 17
Device experience
- Live app
robot/device/live.py - Web server
robot/device/server.py - Browser interface
robot/device/page.html - Video renderer
robot/device/draw.py - Device runtime
robot/device/runtime.py - App entry
robot/app.py - Device settings
robot/config.py
Perception and recognition
- Robot workflow
robot/brain/core.py - Object detection
robot/brain/detect.py
Memory and recall
- Memory operations
robot/brain/memory.py
Voice and models
- Label parsing
robot/brain/labels.py - Embedding and ASR
robot/brain/models.py - Audio capture
robot/device/mic.py
Other components
- User
- Camera
- Microphone
- Qdrant Edge
Connections 21
- User → Live app: uses
- Camera → Live app: supplies frames
- Microphone → Audio capture: supplies audio
- App entry → Robot workflow: creates workflow
robot/app.py - App entry → Live app: starts live mode
robot/app.py - Live app → Web server: starts server
robot/device/live.py - Web server → Browser interface: serves page
robot/device/server.py - Browser interface → Web server: polls and sends controls
robot/device/server.py - Live app → Video renderer: uses renderer
robot/device/live.py - Live app → Device runtime: uses runtime
robot/device/live.py - Live app → Audio capture: uses audio capture
robot/device/live.py - Live app → Robot workflow: dispatches robot actions
- Robot workflow → Object detection: runs detection
robot/brain/core.py - Robot workflow → Embedding and ASR: embeds crops
robot/brain/core.py - Robot workflow → Memory operations: queries memory
robot/brain/core.py - Robot workflow → Label parsing: parses taught names
robot/brain/core.py - Memory operations → Qdrant Edge: stores vectors
README.md - Audio capture → Device settings: reads device setting
robot/device/mic.py - Object detection → Device settings: reads thresholds
robot/brain/detect.py - Embedding and ASR → User: transcribes speech
- Memory operations → User: returns recalled sightings