Build a system that sits between applications and multiple AI providers (OpenAI, Bedrock, Anthropic) and automatically chooses the best and cheapest model for each request. What the system should do (Simple) Take user prompt as input Decide which AI model to use (cheap vs powerful) Send request to that model Cache repeated prompts to save cost Track tokens and cost for every request Show usage and cost in a simple dashboard Key Features (Keep it minimal) Smart Routing → choose cheapest suitable model Caching → avoid paying for same prompt again Cost Tracking → calculate tokens + dollars Basic Dashboard → show usage and savings