Documentation
    Preparing search index...

    Interface OkfEvaluationReport

    Standard qualitative evaluation report returned by a single LLM judge.

    interface OkfEvaluationReport {
        dimensions: OkfDimensionScores;
        evaluatedAt: string;
        grade: "A+" | "A" | "B" | "C" | "D" | "F";
        hallucinations: OkfHallucination[];
        judge: string;
        model: string;
        omissions: string[];
        recommendations: string[];
        score: number;
        strengths: string[];
        summary: string;
        verdict: "WARN" | "PASS" | "FAIL";
    }
    Index

    Properties

    dimensions evaluatedAt grade hallucinations judge model omissions recommendations score strengths summary verdict

    Properties

    dimensions: OkfDimensionScores

    Breakdown of scores across the 5 orthogonal dimensions

    evaluatedAt: string

    ISO timestamp of the evaluation

    grade: "A+" | "A" | "B" | "C" | "D" | "F"

    Academic letter grade based on score

    hallucinations: OkfHallucination[]

    Identified hallucinations with severity grading

    judge: string

    Identifier of the judge ('gemini' | 'chatgpt' | 'claude' | custom)

    model: string

    Model name used for evaluation (e.g. 'gemini-2.5-flash', 'gpt-4o', 'claude-3-5-sonnet')

    omissions: string[]

    Key source concepts or diagnostic items omitted in the extraction

    recommendations: string[]

    Actionable suggestions to improve the extraction prompt or pipeline

    score: number

    Overall weighted qualitative score between 0 and 100

    strengths: string[]

    Strong aspects of the extraction

    summary: string

    Qualitative auditor executive summary

    verdict: "WARN" | "PASS" | "FAIL"

    Automated pipeline verdict

    Settings

    Member Visibility

    On This Page

    Properties