Member Junction
    Preparing search index...

    Interface DriverExecutionResult

    Result from driver execution

    interface DriverExecutionResult {
        actualOutput?: unknown;
        allAgentRunIds?: string[];
        durationMs?: number;
        errorMessage?: string;
        expectedOutput?: unknown;
        failedChecks: number;
        failureClass?: string;
        failureMemo?: string;
        inputData?: unknown;
        oracleResults: OracleResult[];
        outputs?: TestRunOutputItem[];
        passedChecks: number;
        replay?: ReplayTelemetry;
        score: number;
        skippedChecks?: number;
        status: "Passed" | "Failed" | "Skipped" | "Error" | "Timeout";
        targetLogEntityId?: string;
        targetLogId: string;
        targetType: string;
        tier?: string;
        totalChecks: number;
        totalCost?: number;
        totalTurns?: number;
        turnResults?: TurnResult[];
    }
    Index

    Properties

    actualOutput? allAgentRunIds? durationMs? errorMessage? expectedOutput? failedChecks failureClass? failureMemo? inputData? oracleResults outputs? passedChecks replay? score skippedChecks? status targetLogEntityId? targetLogId targetType tier? totalChecks totalCost? totalTurns? turnResults?

    Properties

    actualOutput?: unknown

    Actual output data

    allAgentRunIds?: string[]

    Multi-turn specific: All AgentRun IDs

    durationMs?: number

    Duration in milliseconds

    errorMessage?: string

    Error message if status is Error

    expectedOutput?: unknown

    Expected output data

    failedChecks: number

    Number of checks that failed

    failureClass?: string

    Machine-readable failure classification a driver may compute from its run signals (e.g. Computer Use's nav-loop / stuck-page / app-error / assertion). Free-form so each driver owns its own taxonomy; a retry scheduler keys policy on it. Absent on success.

    failureMemo?: string

    Non-blind retry memo a driver may surface from its engine on a non-passing result — a short "here's what went wrong last time". A retry loop copies it onto PriorAttemptSummary so the next attempt can consume it. Absent on success.

    inputData?: unknown

    Input data used

    oracleResults: OracleResult[]

    Oracle results

    outputs?: TestRunOutputItem[]

    Structured outputs emitted by the driver (screenshots, logs, data, etc.). The engine persists each item as a TestRunOutput entity record.

    passedChecks: number

    Number of checks that passed

    replay?: ReplayTelemetry

    Replay telemetry — present whenever a replay was attempted, so the drift signal survives a green LLM-fallback. Copied to TestRunResult.replay. Absent on pure-LLM runs.

    score: number

    Overall score

    skippedChecks?: number

    Number of checks that were skipped (tier-gated, mutation-gated, or environment gaps). Optional so existing drivers that never skip need no change.

    status: "Passed" | "Failed" | "Skipped" | "Error" | "Timeout"

    Execution status. 'Skipped' means the driver did not execute the test's checks at all (env-gated tier, unreachable dependency) — it must NEVER be reported as 'Passed': a green run that silently didn't execute is indistinguishable from real coverage. The MJ: Test Runs entity's Status value list already includes 'Skipped'.

    targetLogEntityId?: string

    Entity ID identifying the type of target being tested. References Entity.ID (e.g., Entity ID for "MJ: AI Agent Runs"). This is the proper FK reference for entity linkage.

    targetLogId: string

    Target entity ID (final AgentRun ID for single/multi-turn)

    targetType: string

    Optional sub-category or variant label for the test target. Use for ad-hoc labeling or to distinguish test scenarios within the same entity type. Examples: "Summarization", "Classification", "Code Review", "Multi-turn Chat"

    tier?: string

    Execution-tier label a tiered driver reports — e.g. Computer Use's 'replay' / 'replay-with-heal' / 'llm'. Free-form so each driver owns its own tiering. A diverged replay that fell back to the LLM leg reports 'llm' (the tier that produced the result). Absent for single-tier drivers.

    totalChecks: number

    Total number of checks EXECUTED (passed + failed). Skipped checks are counted separately in skippedChecks and never inflate the executed totals.

    totalCost?: number

    Cost in USD

    totalTurns?: number

    Multi-turn specific: Total number of turns

    turnResults?: TurnResult[]

    Multi-turn specific: Results for each turn

    Settings

    Member Visibility

    On This Page

    Properties