Class OkfEntryEvaluator

OkfEntryEvaluator orchestrates autonomous qualitative assessments of extracted OKF documents, supporting standalone Gemini audits and multi-model consensus scoring with ChatGPT and Claude.

Index

Constructors

constructor

new OkfEntryEvaluator(config?: JudgeProviderConfig): OkfEntryEvaluator
Parameters
- config: JudgeProviderConfig = {}
Returns OkfEntryEvaluator
- Defined in judge.ts:251

Properties

`Protected`_config

_config: JudgeProviderConfig

Methods

`Protected`_formatExtractedTarget

_formatExtractedTarget(
extracted: string | OkfDocument | Record<string, unknown>,
): string
Formats an extracted document target into a sanitized inspection string for the judge prompt.
Parameters
- extracted: string | OkfDocument | Record<string, unknown>
Returns string
- Defined in judge.ts:334

`Protected`_normalizePayload

_normalizePayload(
    raw: RawJudgePayload,
    judge: string,
    model: string,
): OkfEvaluationReport
Normalizes raw LLM output into a validated OkfEvaluationReport.
Parameters
- raw: RawJudgePayload
- judge: string
- model: string
Returns OkfEvaluationReport
- Defined in judge.ts:397

buildJudgePrompt

buildJudgePrompt(target: OkfEvaluationTarget): string
Builds the standardized auditor prompt sent to LLM judges.
Parameters
- target: OkfEvaluationTarget
Returns string
- Defined in judge.ts:348

evaluate

evaluate(
    target: OkfEvaluationTarget,
    options?: {
        adapter?: AbstractAiAdapter;
        apiKey?: string;
        judge?: string;
        model?: string;
    },
): Promise<OkfEvaluationReport>
Evaluates an OKF entry with a single specified judge, defaulting to Gemini or first available.
Parameters
- target: OkfEvaluationTarget
- options: { adapter?: AbstractAiAdapter; apiKey?: string; judge?: string; model?: string } = {}
Returns Promise<OkfEvaluationReport>
- Defined in judge.ts:581

evaluateComposite

evaluateComposite(
target: OkfEvaluationTarget,
options?: CompositeJudgeOptions,
): Promise<CompositeEvaluationReport>
Computes a multi-judge composite evaluation, orchestrating parallel audits and consensus synthesis.
Parameters
- target: OkfEvaluationTarget
- options: CompositeJudgeOptions = {}
Returns Promise<CompositeEvaluationReport>
- Defined in judge.ts:613

evaluateWithAdapter

evaluateWithAdapter(
    adapter: AbstractAiAdapter,
    target: OkfEvaluationTarget,
    judgeName?: string,
    modelName?: string,
): Promise<OkfEvaluationReport>
Evaluates an OKF entry using a generic AbstractAiAdapter.
Parameters
- adapter: AbstractAiAdapter
- target: OkfEvaluationTarget
- judgeName: string = 'custom'
- modelName: string = 'custom-model'
Returns Promise<OkfEvaluationReport>
- Defined in judge.ts:465

evaluateWithChatGpt

evaluateWithChatGpt(
target: OkfEvaluationTarget,
options?: { apiKey?: string; model?: string },
): Promise<OkfEvaluationReport>
Evaluates an OKF entry using ChatGPT (OpenAiAdapter).
Parameters
- target: OkfEvaluationTarget
- options: { apiKey?: string; model?: string } = {}
Returns Promise<OkfEvaluationReport>
- Defined in judge.ts:504

evaluateWithClaude

evaluateWithClaude(
target: OkfEvaluationTarget,
options?: { apiKey?: string; model?: string; openRouterKey?: string },
): Promise<OkfEvaluationReport>
Evaluates an OKF entry using Claude (Anthropic API or OpenRouter).
Parameters
- target: OkfEvaluationTarget
- options: { apiKey?: string; model?: string; openRouterKey?: string } = {}
Returns Promise<OkfEvaluationReport>
- Defined in judge.ts:524

evaluateWithGemini

evaluateWithGemini(
target: OkfEvaluationTarget,
options?: { apiKey?: string; model?: string },
): Promise<OkfEvaluationReport>
Evaluates an OKF entry using Gemini (GeminiAdapter).
Parameters
- target: OkfEvaluationTarget
- options: { apiKey?: string; model?: string } = {}
Returns Promise<OkfEvaluationReport>
- Defined in judge.ts:484

`Static`calculateGrade

calculateGrade(score: number): "A+" | "A" | "B" | "C" | "D" | "F"
Computes an academic letter grade from a numerical score (0 to 100).
Parameters
- score: number
Returns "A+" | "A" | "B" | "C" | "D" | "F"
- Defined in judge.ts:258

`Static`calculateVerdict

calculateVerdict(
score: number,
hallucinations: OkfHallucination[],
): "WARN" | "PASS" | "FAIL"
Determines the pipeline verdict based on score and severity of hallucinations. A critical hallucination (such as inventing a species name) triggers an immediate FAIL.
Parameters
- score: number
- hallucinations: OkfHallucination[]
Returns "WARN" | "PASS" | "FAIL"
- Defined in judge.ts:271

`Static`calculateWeightedScore

calculateWeightedScore(dimensions: OkfDimensionScores): number
Computes the weighted composite score from the 5 qualitative dimensions.
Parameters
- dimensions: OkfDimensionScores
Returns number
- Defined in judge.ts:289

`Static`detectAvailableJudges

detectAvailableJudges(config?: JudgeProviderConfig): string[]
Detects which LLM judges can be activated based on available API keys or adapters.
Parameters
- config: JudgeProviderConfig = {}
Returns string[]
- Defined in judge.ts:303

`Static`synthesizeComposite

synthesizeComposite(
reports: OkfEvaluationReport[],
weights?: Record<string, number>,
): CompositeEvaluationReport
Synthesizes multiple individual judge reports into a unified consensus report.
Parameters
- reports: OkfEvaluationReport[]
- Optionalweights: Record<string, number>
Returns CompositeEvaluationReport
- Defined in judge.ts:682

Class OkfEntryEvaluator

Index

Constructors

Properties

Methods

Constructors

constructor

Parameters

Returns OkfEntryEvaluator

Properties

Protected_config

Methods

Protected_formatExtractedTarget

Parameters

Returns string

Protected_normalizePayload

Parameters

Returns OkfEvaluationReport

buildJudgePrompt

Parameters

Returns string

evaluate

Parameters

Returns Promise<OkfEvaluationReport>

evaluateComposite

Parameters

Returns Promise<CompositeEvaluationReport>

evaluateWithAdapter

Parameters

Returns Promise<OkfEvaluationReport>

evaluateWithChatGpt

Parameters

Returns Promise<OkfEvaluationReport>

evaluateWithClaude

Parameters

Returns Promise<OkfEvaluationReport>

evaluateWithGemini

Parameters

Returns Promise<OkfEvaluationReport>

StaticcalculateGrade

Parameters

Returns "A+" | "A" | "B" | "C" | "D" | "F"

StaticcalculateVerdict

Parameters

Returns "WARN" | "PASS" | "FAIL"

StaticcalculateWeightedScore

Parameters

Returns number

StaticdetectAvailableJudges

Parameters

Returns string[]

StaticsynthesizeComposite

Parameters

Returns CompositeEvaluationReport

Settings

On This Page

`Protected`_config

`Protected`_formatExtractedTarget

`Protected`_normalizePayload

`Static`calculateGrade

`Static`calculateVerdict

`Static`calculateWeightedScore

`Static`detectAvailableJudges

`Static`synthesizeComposite