Optimize API

Cost optimization endpoint. Analyzes LLM call patterns and returns model switch recommendations.

GET /api/optimize/recommendations

Returns cost optimization recommendations based on LLM call patterns in the specified period.

Query Parameters

Parameter	Type	Default	Range	Description
agentId	string	—	—	Filter to a specific agent
period	number	7	1–90	Analysis period in days
limit	number	10	1–50	Maximum recommendations to return

Response (200)

{
  "recommendations": [
    {
      "currentModel": "gpt-4o",
      "recommendedModel": "gpt-4o-mini",
      "complexityTier": "simple",
      "currentCostPerCall": 0.0120,
      "recommendedCostPerCall": 0.0003,
      "monthlySavings": 89.20,
      "callVolume": 1203,
      "currentSuccessRate": 0.98,
      "recommendedSuccessRate": 0.97,
      "confidence": "high",
      "agentId": "my-agent"
    },
    {
      "currentModel": "claude-3-opus",
      "recommendedModel": "claude-3-sonnet",
      "complexityTier": "moderate",
      "currentCostPerCall": 0.0450,
      "recommendedCostPerCall": 0.0120,
      "monthlySavings": 53.30,
      "callVolume": 644,
      "currentSuccessRate": 0.95,
      "recommendedSuccessRate": 0.93,
      "confidence": "medium",
      "agentId": "my-agent"
    }
  ],
  "totalPotentialSavings": 142.50,
  "period": 7,
  "analyzedCalls": 3847
}

Response Fields

Field	Type	Description
recommendations	array	Sorted by monthlySavings descending
recommendations[].currentModel	string	Model currently in use
recommendations[].recommendedModel	string	Suggested cheaper alternative
recommendations[].complexityTier	string	simple, moderate, or complex
recommendations[].currentCostPerCall	number	Average cost per call with current model (USD)
recommendations[].recommendedCostPerCall	number	Estimated cost per call with recommended model (USD)
recommendations[].monthlySavings	number	Projected monthly savings (USD)
recommendations[].callVolume	number	Number of calls analyzed
recommendations[].currentSuccessRate	number	Success rate with current model (0–1)
recommendations[].recommendedSuccessRate	number	Estimated success rate with recommended model (0–1)
recommendations[].confidence	string	low, medium, or high
recommendations[].agentId	string	Agent this recommendation applies to
totalPotentialSavings	number	Sum of all recommendation savings (USD/month)
period	number	Analysis period used (days)
analyzedCalls	number	Total LLM calls analyzed

Errors

Status	Condition
400	Invalid period or limit parameter