Get Model Analytics

Get comprehensive analytics for a specific model including quality scores, sentiment analysis, safety metrics, LLM-generated recommendations, and time-series data.

Requires Growth+ plan or higher.

Rate limited based on your plan’s maxAnalysesPerHour.

GET /models/{modelId}/analytics

cURL

cURL

curl --request GET \
  --url https://www.greenflash.ai/api/v1/models/{modelId}/analytics \
  --header 'Authorization: Bearer <token>'

200

{
  "avgConversationQualityIndex": 7.8,
  "avgSatisfactionScore": 0.82,
  "avgGrowthScore": 0.65,
  "avgFrictionScore": 0.88,
  "avgSafetyScore": 0.94,
  "averageUserSentiment": {
    "label": "positive",
    "score": 0.28
  },
  "averageChangeInUserSentiment": {
    "label": "stable",
    "score": 0.03
  },
  "averageFrustration": {
    "label": "low",
    "score": 0.18
  },
  "averageStruggle": {
    "label": "low",
    "score": 0.14
  },
  "averageCommercialIntent": {
    "label": "moderate",
    "score": 0.42
  },
  "hallucinationCount": {
    "total": 8,
    "percentage": 0.9
  },
  "jailbreakCount": {
    "total": 2,
    "percentage": 0.2
  },
  "userBiasCount": {
    "total": 1,
    "percentage": 0.1
  },
  "modelBiasCount": {
    "total": 3,
    "percentage": 0.3
  },
  "userToxicityCount": {
    "total": 4,
    "percentage": 0.4
  },
  "modelToxicityCount": {
    "total": 0,
    "percentage": 0
  },
  "recommendations": {
    "summary": "GPT-4o is performing well overall. Hallucination rate is slightly elevated for refund policy topics.",
    "healthLevel": "healthy",
    "strengths": [
      "High satisfaction scores across all products",
      "Excellent safety compliance",
      "Low friction in multi-turn conversations"
    ],
    "recommendations": [
      {
        "title": "Reduce hallucination in billing topics",
        "action": "Add explicit billing policy context to the system prompt",
        "rationale": {
          "whyThisMatters": "Billing hallucinations erode customer trust.",
          "observedPattern": "65% of hallucinations occur in refund/billing conversations.",
          "tradeoff": "Longer system prompt may slightly increase latency."
        },
        "evidence": {
          "exampleSnippets": [
            "Bot stated \"refunds are processed in 24 hours\" when policy says 5-7 business days."
          ]
        },
        "expectedImpact": [
          "Reduce hallucination rate by ~40%",
          "Improve billing conversation CQI from 6.2 to 7.5+"
        ],
        "priority": "high",
        "category": "prompt_optimization_link"
      }
    ],
    "detectedIssues": [
      {
        "id": "hallucination-billing",
        "severity": "warning",
        "title": "Elevated hallucination rate in billing-related conversations"
      }
    ]
  },
  "conversationCount": 4520,
  "messageCount": 31640,
  "averageConversationRating": 4.2,
  "dataPoints": [
    {
      "date": "2025-10-01T00:00:00Z",
      "avgConversationQualityIndex": 7.6,
      "avgSatisfactionScore": 0.8,
      "avgGrowthScore": 0.63,
      "avgFrictionScore": 0.86,
      "avgSafetyScore": 0.93,
      "conversationCount": 152,
      "messageCount": 1064
    },
    {
      "date": "2025-10-08T00:00:00Z",
      "avgConversationQualityIndex": 7.9,
      "avgSatisfactionScore": 0.83,
      "avgGrowthScore": 0.66,
      "avgFrictionScore": 0.89,
      "avgSafetyScore": 0.95,
      "conversationCount": 168,
      "messageCount": 1176
    }
  ]
}

Authorizations

Authorization · string · header · required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Path Parameters

modelId · string<uuid> · required

The model ID to get analytics for

Query Parameters

period · enum<string> · default: 30d

Time period for analytics data: 7d, 30d (default), or 90d.

Available options: 7d, 30d, 90d

Response

200

application/json

Model analytics retrieved successfully

avgConversationQualityIndex · number | null · required

Average conversation quality index.

avgSatisfactionScore · number | null · required

Average satisfaction score.

avgGrowthScore · number | null · required

Average growth score.

avgFrictionScore · number | null · required

Average friction score.

avgSafetyScore · number | null · required

Average safety score.

averageUserSentiment · object · required

Average user sentiment.

Show child attributes

averageChangeInUserSentiment · object · required

Average change in user sentiment.

Show child attributes

averageFrustration · object · required

Average frustration level.

Show child attributes

averageStruggle · object · required

Average struggle level.

Show child attributes

averageCommercialIntent · object · required

Average commercial intent.

Show child attributes

hallucinationCount · object · required

Hallucination detection counts.

Show child attributes

jailbreakCount · object · required

Jailbreak detection counts.

Show child attributes

userBiasCount · object · required

User bias detection counts.

Show child attributes

modelBiasCount · object · required

Model bias detection counts.

Show child attributes

userToxicityCount · object · required

User toxicity detection counts.

Show child attributes

modelToxicityCount · object · required

Model toxicity detection counts.

Show child attributes

recommendations · object · required

LLM-generated recommendations and health assessment.

Show child attributes

conversationCount · number · required

Total conversation count.

messageCount · number · required

Total message count.

averageConversationRating · number | null · required

Average conversation rating.

dataPoints · object[] · required

Time-series data points for the requested period.

Show child attributes