AI编码Agent智能模型路由配置生成器

You are an expert in LLM cost optimization and model routing. Help me design an intelligent model routing configuration for my AI coding workflow.

My Setup

• Monthly budget: [BUDGET, e.g. $100]
• Primary use cases: [e.g. code generation, debugging, code review, documentation]
• Available models/providers: [e.g. OpenAI, Anthropic, local models, Gemini]
• Current monthly spend: [CURRENT_SPEND]

Design a Routing Strategy

1. Task Classification Rules

Define clear rules to classify incoming requests into complexity tiers:

• Tier 1 (Simple): Examples and routing criteria
• Tier 2 (Medium): Examples and routing criteria
• Tier 3 (Complex): Examples and routing criteria
• Tier 4 (Expert): Examples and routing criteria

2. Model Assignment

For each tier, specify:

• Primary model (with rationale)
• Fallback model
• Max tokens / context window needs
• Expected cost per request

3. Classification Heuristics

Provide concrete heuristics based on:

• Input token count
• Presence of code blocks
• Question complexity indicators
• Language/framework specificity
• Whether it requires reasoning chains

4. Cost Projection

• Estimated monthly cost with routing vs. single model
• Savings percentage
• Quality trade-off analysis

5. Configuration Output

Generate a ready-to-use JSON/YAML configuration for a model router (compatible with LiteLLM, Manifest, or similar tools).

6. Monitoring Recommendations

• Key metrics to track
• When to adjust routing rules
• A/B testing strategy for quality validation