AI Agents
Multi-agent system running 24/7 — click any agent to see its architecture, data sources, and source code.
⚙️ System Overview
Agents run in isolated environments on the same host, sharing a Redis message bus. Each agent publishes structured logs to a central store. The scheduler (cron) triggers agents on fixed intervals — Market Agent every 5 min, Report Agent weekly on Fridays. All agents share access to the Docker-mounted pages directory for output persistence.
Market Analysis Agent
Architecture
Data Sources
Tickers tracked: TSLA, SPCX, FLY, RKLB, IONQ, RGTI, QBTS, QUBT
Data points per ticker: Price, change %, 52W range, volume, trailing P/E, float shares, outstanding shares
MAVI formula: earnings_yield × float_ratio × log₁₀(float_shares × price)
Signal Aggregation Agent
Architecture
Data Sources
Scoring methodology: Each sub-agent contributes a normalized score (0–10). The aggregator applies weights: MAVI=0.4, Price Trend=0.35, Volume=0.25. A weighted score ≥ 6.5 triggers a BUY signal, ≤ 3.5 triggers SELL.
Report Generation Agent
Architecture
Data Sources
Output files: /stocks/data/current.json (latest snapshot) + /stocks/data/history.json (52-week history) + /stocks.html (main page update)
Execution Agent
Architecture
Data Sources
Thresholds: Consensus ≥ 7.5 → STRONG BUY alert · Consensus ≤ 2.0 → STRONG SELL alert · Others logged as info
Backup Agent
Architecture
Data Sources
Retention policy: 30 most recent backups kept; older snapshots auto-deleted each run. Manual backups available anytime via backup-sites.
Recovery Agent
Architecture
Data Sources
Recovery playbook: If a site breaks — check docker ps -a for exit code 127 (nginx config mounted as directory). Fix: remove empty dir, restore config file, docker start. If files are missing: deploy-sites from workspace. If you need an older version: restore-site <timestamp>.