EvaluateStatistics interface

LLM response time statistics emitted after each evaluation response.

Signature:

export interface EvaluateStatistics

Properties

Property	Type	Description
averageCallIntervalMs	number	Observed spacing between LLM calls; compare with the configured analysisIntervalSecs to see backpressure.
averageResponseTimeMs	number
cacheStatus	string	Provider cache lifecycle: "created" or "skipped:" for the Gemini content cache, "automatic" for OpenAI, "" when no cache is in play.
callsSkippedByGate	number	Interval ticks where a call would have been made but the evaluation gate was closed (the node emits onAnalysis instead on those ticks).
consecutiveStateFailures	number	State-loop health: consecutive resets on every successfully applied _update_state. See stateFailure in the node settings for escalation.
framesDroppedPendingRequest	number	Sampled frames discarded because an LLM request was still in flight — a slow provider silently lowers the effective sample rate.
lastCachedTokens	number	Input tokens served from the provider’s prompt cache on the last response (Gemini explicit content cache / OpenAI automatic caching).
lastInputTokens	number
lastOutputTokens	number
lastResponseTimeMs	number
sampleCount	number
totalCachedTokens	number
totalInputTokens	number	Cumulative token counts. Together with the provider’s published per-token prices these give the running cost of the session (cached input tokens are usually billed at a steep discount — subtract totalCachedTokens from totalInputTokens for the full-price share).
totalOutputTokens	number
totalStateFailures	number

Property

Type

Description

averageCallIntervalMs

number

Observed spacing between LLM calls; compare with the configured analysisIntervalSecs to see backpressure.

averageResponseTimeMs

number

cacheStatus

string

Provider cache lifecycle: "created" or "skipped:" for the Gemini content cache, "automatic" for OpenAI, "" when no cache is in play.

callsSkippedByGate

number

Interval ticks where a call would have been made but the evaluation gate was closed (the node emits onAnalysis instead on those ticks).

consecutiveStateFailures

number

State-loop health: consecutive resets on every successfully applied _update_state. See stateFailure in the node settings for escalation.

framesDroppedPendingRequest

number

Sampled frames discarded because an LLM request was still in flight — a slow provider silently lowers the effective sample rate.

lastCachedTokens

number

Input tokens served from the provider’s prompt cache on the last response (Gemini explicit content cache / OpenAI automatic caching).

number

number

number

number

number

number

Cumulative token counts. Together with the provider’s published per-token prices these give the running cost of the session (cached input tokens are usually billed at a steep discount — subtract totalCachedTokens from totalInputTokens for the full-price share).

totalOutputTokens

number

totalStateFailures

number