personal-brand-agent

The LinkedIn Super Agent is a sophisticated autonomous system designed to build and maintain a professional personal brand on LinkedIn. By leveraging a multi-agent orchestration architecture via LangGraph, the system executes a full-stack content pipeline from trend discovery to publication and performance analytics. It utilizes state-of-the-art LLMs, specifically Claude Opus 4.5 in thinking mode, to ensure high-quality, nuanced content generation that avoids generic AI patterns.

• Seven specialized agents covering the entire lifecycle: Trend Scout, Analyzer, Writer, Humanizer, Visual Creator, Photo Selector, and Quality Control.

• Self-improvement engine that generates, tests, and hot-reloads its own Python code at runtime to adapt to changing platform dynamics.

• Deterministic LangGraph state machine with conditional routing, retry loops, and fail-fast error handling.

• Advanced author voice profiling that analyzes historical posts to maintain a consistent professional tone.

• Configurable autonomy levels (1-4), supporting everything from human-in-the-loop approval via Telegram to fully autonomous publishing.

• Automated analytics collection at specific intervals (T+15min to T+24h) for continuous benchmarking.

• Ensure all environment variables (ANTHROPIC_API_KEY, SUPABASE_URL, etc.) are correctly configured in your .env file.

• Run the database migration script before initial deployment to ensure schema compatibility.

• Use the provided scheduling daemon for hands-off operation or trigger the orchestrator manually for controlled generation.

• Monitor the Telegram bot interface for human-in-the-loop interactions when autonomy levels require approval.

• The system integrates with Perplexity API for research, ArXiv for academic context, and LaoZhang.ai for image generation.