Omi Open-Source Second Brain Platform

Omi Triangle v2 is an upgraded hardware version of the Omi Triangle v1. Key improvements in v2 include:

• Added Memory: Support for external storage via a Micro SD card.
• Speaker: Integrated audio output capabilities.
• Configurable Momentary Button: A tactile button for user interaction.

Note: This is currently a developer version. Some new hardware functionality may not be fully supported by the existing firmware. If you are using a device assembled before August 23, 2024, you must perform a speaker function correction to ensure proper operation.

This audit compares the capabilities of the File Index, Knowledge Graph (KG), and Memory Graph features between macOS and Windows implementations in OMI.

Key functional areas include:

• File Indexing: Scanning roots, directory-skip policies, incremental scans, and file-type categorization.
• Knowledge Graph (KG): Local storage schemas (nodes/edges), backend fetching/rebuilding, and client-side scoping.
• Onboarding & AI: AI-driven file exploration for entity extraction (creating a 'digital profile') and chat agent context enrichment.
• Memory Graph Visualization: 3D/2D physics models and interactive UI (drag, rotate, pan, zoom).

Note: While many features are present on Windows, some (like incremental background refreshes and AI-driven onboarding exploration) are currently absent or weaker compared to the macOS baseline.

The Omi Python SDK is a pip-installable library designed to interface with Omi wearable devices. It provides capabilities for:

• Bluetooth Connection: Connecting to Omi hardware devices.
• Opus Decoding: Decoding Opus-encoded audio streams into PCM format.
• Real-time Transcription: Leveraging Deepgram for real-time Speech-to-Text (STT).

The Omi firmware is the core software for the Omi AI wearable device. It is built on the Zephyr RTOS and manages critical hardware functions including:

• Audio Capture: Microphone input and audio buffering.
• Codec: Processing raw audio data.
• Transport: Bluetooth connectivity and audio streaming.
• Storage: SD card operations and audio file management.
• Battery & LED Control: Power management and visual status feedback.

Note: Due to its complexity, this firmware cannot be built using the Arduino IDE; it requires a professional toolchain (typically involving CMake and the nRF Connect extension).

Omi DevKit 2 is an upgraded wearable in a necklace form factor. Key hardware specifications and features include:

• Processor: Xiao nRF52840
• Storage: 8GB onboard storage, which enables Standalone Recording Mode (recording without a phone connection).
• Audio: Built-in speaker for audio feedback and notifications.
• Interaction: A programmable button that replaces the traditional on/off switch.

Note: While DevKit 2 supports standalone recording, you must still connect the device to the Omi app for transcription processing.

The Omi backend uses a Prometheus, Grafana, and Loki observability stack running on Google Kubernetes Engine (GKE). This stack provides metrics collection, log aggregation, and visualization for all backend services.

Core Components:

• Prometheus: Metrics collection with 10-day retention and 50Gi storage.
• Grafana: Visualization and alerting. Access via monitor.omi.me (prod) or monitor.omiapi.com (dev).
• Loki: Log aggregation using a GCS backend with 15-day retention.
• Alloy: A DaemonSet used for collecting pod logs via the Kubernetes API.
• Prometheus Adapter: Translates Prometheus metrics into Kubernetes custom/external metrics APIs to enable Horizontal Pod Autoscaling (HPA).
• Stackdriver Exporter: Bridges GCP Load Balancer metrics into Prometheus.