Source: Finnhub

Model: OpenAI gpt-4o-mini preferred (fast + lower cost), fallback Ollama