Skip to main content

CROW Internal Documentation

Welcome to the internal documentation for CROW (Cognitive Reasoning Observation Watcher) by B3.

About CROW​

CROW is a unified customer interaction intelligence platform designed to address a fundamental challenge: customer behavior signals generated across digital storefronts, physical retail, and social media remain fragmented within isolated systems. This fragmentation prevents organizations from developing a unified understanding of shopper behavior.

The Problem​

Organizations face three interconnected challenges:

  1. Collection: Raw signals are difficult to capture comprehensively across web, in-store, and social channels
  2. Interpretation: Captured data lacks contextual understanding to transform into meaningful insights
  3. Synthesis: Insights cannot be combined into actionable patterns across channels

The Solution​

CROW consolidates multimodal interaction signals into a single, searchable natural language data layer:

  • Web Analytics: Lightweight JavaScript SDK captures user interactions
  • In-Store: CCTV footage processed through real-time video streaming with Gemini Live
  • Social Media: Scheduled discovery agents collect public brand mentions

The platform transforms raw behavioral signals into contextually enriched, human-readable interaction records. Sales, marketing, product, and operations teams can access unified customer insights through dashboards and APIs without manual cross-channel data reconciliation.

Three Core Components​

CROW operates through three main data collection components, deployed in phases:

1. Web Component (Phase 1 - Current Focus)​

The Web component provides an SDK that integrates with brand-owned applications (web, mobile via React Native, and desktop via Electron). This lightweight SDK captures user interactions, events, and behaviors across any user interface with a DOM setup. The SDK sends interaction data to our platform where it's processed and analyzed to generate insights.

2. Social Component (Phase 2)​

The Social component performs intelligent web searching and extraction of social media data. It analyzes company social media posts, comments, and engagement across various platforms to gather insights at a deep level, tracking public sentiment and engagement patterns.

3. CCTV Component (Phase 3)​

The CCTV component represents our most advanced offering. It streams CCTV footage from retail stores through Cloudflare's real-time serverless SFU using WebRTC. The streams are converted via WebRTC-to-WebSocket adapters and fed to Gemini Live for real-time analysis, providing descriptions of customer movements, behaviors, and interactions within physical spaces.

Key Differentiators​

  • Centralized Platform: Single source of truth for all interaction data across channels
  • Contextual Analysis: Not just raw data, but interactions analyzed with full business context
  • Pattern Recognition: Identifies trends and patterns across all collected data
  • Universal Value: A data platform that can be utilized by anyone within an organization, with unlimited potential applications

Documentation Overview​

Architecture​

Features​

Reference​

Getting Started​

Use the sidebar to navigate through documentation sections.

Contributing​

To update documentation:

  1. Edit markdown files in the docs/ directory
  2. The site rebuilds automatically with your changes
  3. All documentation is version-controlled