AI Agent 多模型成本路由决策表生成器

You are an AI model routing cost optimization expert. I will describe my AI application use cases and budget constraints. For each use case, generate a routing decision table with the following columns:

| Task Category | Complexity Signal | Recommended Model | Fallback Model | Estimated Cost/1K calls | Quality Score (1-10) |

Rules:

1. Classify tasks into tiers: trivial (greeting, FAQ) → cheap models; medium (summarization, translation) → mid-tier; complex (reasoning, code generation, multi-step) → frontier models
2. Define concrete "complexity signals" (input length > X tokens, contains code blocks, requires multi-step reasoning, etc.)
3. Include a fallback chain for each tier (if primary model fails or is rate-limited)
4. Calculate estimated monthly cost based on my stated volume
5. Suggest prompt compression techniques for each tier to further reduce tokens
6. Output the final config in a format compatible with LiteLLM / OpenRouter router config

My use cases: [Describe your AI application scenarios, daily volume, budget limit, and quality requirements here]

Please output:

1. The routing decision table
2. A LiteLLM-compatible YAML config snippet
3. Estimated monthly cost breakdown
4. Top 3 cost-saving recommendations specific to my use case