Get cross-version analytics comparison insights
Computes deterministic red flags / proactive insights across versions (metric degradation, metric improvement, cost outlier, latency outlier), plus a fail-open LLM-generated narrative summary. When versionIds is omitted, all versions present in the analytics data are compared. See Platform Overview.
Authorizations
API key authorization. Pass your API key in the Authorization header as a Bearer token. Both new (gsk_*) and legacy (gsk-) API keys are accepted, e.g. Authorization: Bearer gsk_... or Authorization: Bearer gsk-....
Query Parameters
Product ID (required)
Version IDs to compare. When omitted, all versions present in the analytics data are compared.
Filter by test IDs
Filter by metric IDs
Filter by language codes
Start date for filtering
End date for filtering
When true (default), a fail-open LLM-generated narrative summary is computed and returned alongside the deterministic flags. When false, the narrative LLM call is skipped entirely and the response carries narrative: null and narrativeGeneratedByAi: false; the deterministic flags are unchanged. Machine consumers that produce their own prose should pass false to avoid the wasted LLM call.
When set, returns only production (true) or only development (false) data. Omit to include both.
Response
Cross-version analytics comparison insights computed successfully
Cross-version analytics comparison insights: deterministic flags plus a fail-open LLM-generated narrative summary (the narrative is omitted when the request sets includeNarrative=false).
"prod_123"
all_good, flags_present "flags_present"
True when the narrative was successfully produced by the LLM; false otherwise (including when includeNarrative=false).
LLM-generated narrative summary, or null when generation was skipped (includeNarrative=false or no flags) or failed.
"Faithfulness regressed from v2 to v3 while latency on v3 is 1.6× the cheapest version."