architecture
multi-agent: 10 core agents + 9 domain specialists
core agents (10)
model-selector, prompt-enhancer, clarification, simple-responder, executor, specialist-selector, planner, react-worker, validator, debater
specialists (9)
financial, research, data, strategy, clinical, biotech-bd, legal-ip, capital-markets, scientific
tools (6)
google-search, web-scrape, calculator, code-interpreter, data-analysis, financial-calc
features
adaptive-model-selection, fast-path-routing, clarification-system, quality-validation, adversarial-debate, conversation-memory
performance
simple: 2-3s | complex: 15-30s | quality: 75-85 (stricter validation)