OmniWarRoom AI is an autonomous competitive intelligence system built on a LangGraph multi-agent swarm. It continuously monitors competitor websites, bypasses bot-protection layers (DataDome, Akamai, Cloudflare) via Bright Data Web Unlocker, and cross-references live Google SERP data to detect pricing changes, marketing campaigns, and product anomalies in real time. The swarm runs four specialized agents in sequence: a Scout (Groq LLaMA 3.1-8B) that collects live web data via Bright Data MCP tools using streamable_http transport, an Analyst (Gemini 2.0 Flash with Pydantic schema enforcement) that scores the competitive threat from 1 to 5 and identifies structured signals with category and severity, a Tactician (Groq LLaMA 3.3-70B) that activates when the threat threshold is reached and generates 3-5 ROI-framed counter-actions, and a Persistence node that indexes all intelligence into a Cognee knowledge graph (LanceDB + NetworkX) for cross-cycle memory. The system runs as a fully autonomous worker loop decoupled from the UI, with parallel URL scanning via asyncio.gather(), atomic brain-state writes via os.replace(), and a multi-channel alert router that dispatches branded HTML emails, Triggerware GTM webhooks, and voice briefings simultaneously on critical threats. A Streamlit dashboard provides a live SVG agent visualization, voice command input via Speechmatics REST, real-time Plotly threat charts, and a human-in-the-loop approval gate for every generated counter-strategy.
Category tags: